Pathos AI
精准医疗前沿的 AI 驱动肿瘤药物开发
Pathos AI 已搭起可信的 AI 药物开发平台,既有强势药企合作,也有多资产肿瘤管线;但临床和监管去风险仍处早期,关键人物与集中度风险不小
封面要素
公司概况
Pathos AI 是一家总部位于 Chicago 的临床阶段、AI 赋能生物技术公司,用多模态肿瘤数据和自研 AI 重构药物开发。公司由 Eric Lefkofsky 和 Ryan Fukushima 于 2022 年创立,累计融资约 $467M,其中包括 2025 年 5 月以约 $1.6B 估值完成的 $365M Series D。其 PathOS 平台(Scout、Sprint、Foundry) 识别低估临床资产、设计由生物标志物驱动的试验,并与 AstraZeneca 和 Tempus 合作构建肿瘤基础模型。领先资产 pocenbrodib(CBP/p300 抑制剂)处于 1b/2a 期临床试验。
- 成立时间
- 2022-01-01
- 创始人
- Eric Lefkofsky, Ryan Fukushima
- 创立地点
- Chicago, IL
- 总部
- Chicago, IL
- 产品
- PathOS 平台包括 Scout(资产识别)、Sprint(试验设计 / CDx)和 Foundry(AI/ML 模型构建);临床管线包括 pocenbrodib(CBP/p300 抑制剂)、MET 抑制剂和 PRMT5 抑制剂
- 客户
- 寻求 AI 加速肿瘤药物开发的制药与生物技术公司;内部肿瘤治疗管线
- 商业模式
- 双轨模式:一边推进内部药物开发管线,未来可能靠里程碑、版税、许可和资产价值兑现获益;一边与 AstraZeneca、Novo Nordisk、Tempus 开展战略研发合作,随时间形成协作经济收益
- 阶段
- Series D
- 融资情况
- $365M Series D(2025 年 5 月,估值约 $1.6B);累计融资约 $467M
执行摘要
主要优势
- 200+ PB 多模态肿瘤数据集,加上 Tempus AI 和 AstraZeneca 数据合作,撑起 AI 肿瘤药物开发的数据护城河
- Iker Huerga 与 Eric Lefkofsky 领衔,团队经验充足;NEA 及 11 家以上 Series D 投资方背书,显示机构信心
- pocenbrodib、MET 抑制剂、PRMT5 抑制剂及来自 Rain 的资产构成多资产肿瘤管线,相比典型临床期 biotech,单一资产风险更低
主要风险
- 临床执行风险:pocenbrodib Phase 1b/2a 尚未披露疗效数据,AI 筛选资产仍要拿出传统临床证据
- 关键人物和关联方集中度高:平台高度依赖 Tempus 相关数据入口,也存在管理层重叠
- 对一家尚无收入、尚未进入 Phase 2 的平台给出约 $1.6B 估值,相对已披露进展,退出门槛很高
未决问题
- Series D 投资方身份和条款大多未披露;股权结构表和优先股堆叠未知
- 公开渠道没有审计财务报表、烧钱速度、现金余额或现金跑道披露
- pocenbrodib Phase 1b/2a 临床疗效数据尚未公布,公开的 AI 到获批路径记录也尚未出现
目录
01公司概况
1.1 身份、总部与阶段
Pathos AI, Inc. 是一家临床阶段人工智能与生物技术公司,注册地为 Delaware,总部位于 600 W. Chicago Ave., Suite 510, Chicago, Illinois 60654。公司的 SEC CIK 为 0001967854,IRS EIN 为 852945509。Pathos 成立于 2022 年,交汇点是技术、肿瘤学和亲历癌症的经验。官方网站为 www.pathos.com,财政年度截至 12 月 31 日。 公司给出的使命是借助自研患者数据和 AI 改造药物开发,大幅提高试验成功率,把新疗法带给传统治疗失败的患者。Pathos 将自己定位为 AI 赋能生物技术公司,靠 AI 技术、多模态真实世界数据和患者来源生物模型重构药物开发。截至本次报告生成日,Pathos 仍处临床阶段,领先项目 pocenbrodib 正在开展 1b/2a 期临床试验,管线另有三个肿瘤资产。公司未公开员工数。 Pathos 将竞争差异化定义为两点叠加:最大规模的自研多模态肿瘤数据集(声称超过 200 PB,并与患者结局关联),以及持续学习的 AI 平台。这让它区别于纯计算型 AI 公司和传统药企收购方。公司强调,每个完成的试验都会反哺下一个试验,形成学习飞轮,持续放大效率和证据质量。[CO001, CO002, CO003, CO004, CO035, CO037]
| 指标 | 数值 / 状态 | 日期 | 置信度 | 缺口 / 注意事项 |
|---|---|---|---|---|
| 总部 | 总部地址:600 W. Chicago Ave., Suite 510, Chicago, IL 60654 | 2025-05-01 | 高 | SEC Form D 确认 |
| 成立年份 | 2022 | 2022-01-01 | 中 | 具体月份未公开披露 |
| 注册州 | 特拉华州 | 2023-03-02 | 高 | SEC Form D 确认 |
| 阶段 | 临床阶段(1b/2a 期进行中) | 2025-03-20 | 高 | pocenbrodib 试验于 2025 年 3 月启动 |
| 最新投后估值(USD) | ~$1.6 billion(Series D) | 2025-05-15 | 中 | 公司口径;无独立验证 |
| 累计融资(估计) | ~$467 million | 2025-05-15 | 中 | 披露轮次金额加总;Series A 前细节不可得 |
| Series D Form D 发行额(USD) | $399,999,933 | 2025-05-01 | 高 | 直接来自 SEC Form D 文件 |
| Series D Form D 已售金额(USD) | $282,999,950 | 2025-05-01 | 高 | 直接来自 SEC Form D;余额可能反映后续交割 |
| 员工数 | 未公开披露 | 2026-05-25 | 低 | 公开文件或新闻稿没有员工数数据 |
| 主要临床资产 | Pocenbrodib(CBP/P300 抑制剂,mCRPC) | 2025-03-20 | 高 | 进行中的 1b/2a 期试验 NCT06785636 |
估值和累计融资来自公司披露或按已披露轮次金额计算;公开来源无法取得员工数。Form D 金额反映 SEC 文件日期,可能不同于新闻稿报道的最终轮次交割金额。
[CO001, CO002, CO003, CO010, CO012, CO013]1.2 创始人、领导层与治理
Pathos AI 由 Eric Lefkofsky 和 Ryan Fukushima 于 2022 年共同创立。Eric Lefkofsky 同时是 Tempus AI, Inc.(Nasdaq: TEM)的创始人兼 CEO;Tempus 是一家专注 AI 赋能精准医疗的健康科技公司。该公开关系解释了 Pathos 为何能深度访问 Tempus 数据集。Ryan Fukushima 曾担任公司创始及临时 CEO,任期至 2025 年初。 Iker Huerga 于 2025 年 5 月加入 Pathos AI,担任 CEO 和董事会成员。Huerga 是癌症幸存者,也有生物技术行业资历;此前最近一职是 AstraZeneca 肿瘤研发首席数据科学家,直接贴合 Pathos 正在推进的 AstraZeneca 合作。Dr. Jens Renstrup 担任首席医疗官,并主导 pocenbrodib 的临床策略。公开资料列 New Enterprise Associates(NEA)联席 CEO Mohamad Makhzoumi 为董事会成员。 GenomeWeb 曾发表一篇文章,将 Pathos 初称为 Tempus 分拆公司,后续予以更正;更正说明 Pathos 由 Tempus 高管创立,但法律上独立、运营上独立。这一事件说明,Pathos–Tempus 关系会受到外部审视,描述时必须谨慎:Pathos 不是 Tempus 实体,但人员和数据重叠对独立性及利益冲突尽调有实质意义。 关键人依赖偏高:Fukushima 到 Huerga 的 CEO 交接发生在 2025 年 5 月,正好与 Series D 完成重合,暗示这是一次有意安排的融资前管理层升级。除 Makhzoumi 外,完整董事会构成尚未公开。[CO004, CO005, CO006, CO007, CO008, CO009]
| 人物 | 职务 | 背景 | 创始人-市场匹配 / 覆盖 | 关键人物依赖 |
|---|---|---|---|---|
| Iker Huerga | CEO 兼董事会成员(2025 年 5 月任命) | 癌症幸存者;曾任 AstraZeneca 肿瘤研发首席数据科学家 | 具备直接的肿瘤 AI 与 AZ 合作经验;生物科技运营老将 | 高 — 任命后唯一公开高管面孔;此前无 Pathos 任职经历 |
| Ryan Fukushima | 联合创始人兼创始 CEO(交接后角色不明确) | 2022 年与 Eric Lefkofsky 共同创立 Pathos AI;主导 Series A 至 Series C | 搭建创始团队、平台和首批资产收购 | 中 — CEO 交接带来战略连续性风险 |
| Eric Lefkofsky | 联合创始人兼董事会成员(非执行) | Tempus AI(Nasdaq: TEM)创始人兼 CEO;连续创业者(Groupon、Mediaocean) | 借 Tempus AI 获得数据访问与战略网络;有资本形成经验 | 高 — Tempus 数据合作与 AstraZeneca 关系流部分来自 Lefkofsky 网络 |
| Dr. Jens Renstrup | 首席医疗官 | 临床肿瘤学经验;引用 COURAGE 研究数据主导 pocenbrodib 试验策略 | 主导核心项目临床开发 | 中 — 医学策略集中在 CMO 角色 |
| Mohamad Makhzoumi | 董事会成员(通过 NEA) | New Enterprise Associates 联席 CEO;主导 Series C 投资 | 治理监督与资本配置;生命科学投资经验 | 低 — 董事会成员,不是运营者 |
角色和背景来自公司新闻稿及官网;Fukushima 交接后的角色未获公开来源确认。除 Makhzoumi 外,完整董事会构成未公开披露。
[CO005, CO006, CO007, CO008, CO009, CO017]1.3 资本历程、估值与投资者版图
Pathos AI 在四次已披露融资事件中累计融资约 $467M。2023 年 3 月 Form D(SEC filing number 021-474736)下首次募集 $40M,用于平台开发和首个资产收购。随后公司在 2024 年 10 月完成超额认购的 $62M Series C,由 New Enterprise Associates 领投,投后估值 $600M;当时报道的累计融资为 $102M,印证 Series C 前的 $40M。Series D 于 2025 年 5 月 15 日宣布完成,融资 $365M,将投后估值推至约 $1.6B。2025 年 5 月 1 日 Form D 显示发行总额 $399,999,933,截至申报日已向 11 名投资者售出 $282,999,950;考虑后续交割,这与公告金额一致。 Series D 投资者身份尚未公开披露。Pathos 确认本轮包括老股东和新投资者。Series C 投资者基础(NEA、Revolution Growth、Lightbank、Builders VC)是目前记录最完整的一组。2025 年 4 月基础模型合作协议下向 Tempus AI 支付的 $200M,是一笔重要的关联方商业交易,需要单独尽调——资金从 Pathos 流向作为交易对手的 Tempus,而非股权付款。 未公开披露债务融资、老股交易或信贷额度。Pathos 仍为私有公司,且因不披露财务报表,可归入私有未披露类别。[CO010, CO011, CO012, CO013, CO014, CO015]
| 利益相关方 | 角色 | 轮次 / 事件 | 经济 / 控制重要性 | 尽调问题 |
|---|---|---|---|---|
| New Enterprise Associates(NEA,投资方) | 领投方;董事会席位 | 领投 $62M Series C(2024 年 10 月) | 公开记录最充分的投资者;Makhzoumi 在董事会 | 确认董事席位条款、Series D 中的按比例跟投权、投票权 |
| Revolution Growth | 共同投资者 | Series C 参与方 | 与 NEA 并列的早期机构支持者 | 确认持股规模及任何治理权利 |
| Lightbank | 早期投资者 | Seed/Series A 至 Series C | 早期轮次与 Builders VC 共同投资 | 说明初始出资规模和稀释路径 |
| Builders VC | 早期投资者 | Seed/Series A 至 Series C | 早期轮次与 Lightbank 共同投资 | 说明持股和任何顾问安排 |
| Series D 投资者(未具名组合) | 新老投资者 | $365M Series D(2025 年 5 月) | 最大资本提供方;身份未披露 | 识别 Series D 领投方及是否有董事会席位 |
| Tempus AI (Nasdaq: TEM) | 战略伙伴和数据许可方 | 2025 年 4 月合作;向 Tempus 支付 $200M 许可费 | 关键数据与模型开发伙伴;Eric Lefkofsky 是 Tempus CEO,也是 Pathos 联合创始人 | 评估利益冲突治理;审查许可条款和排他性条款 |
| AstraZeneca (AZN) | 战略伙伴和模型联合开发方 | 2025 年 4 月多年合作 | 验证 AI 平台可被生物制药采用;共同拥有产出的基础模型 | 确认期限长度、IP 权属拆分和商业化权利 |
Series D 投资者身份未公开披露。利益相关方关系来自 Pathos 官方新闻稿和 SEC Form D 文件。Tempus 与 AstraZeneca 是商业伙伴,不是股权投资者,但对商业模式很关键。
[CO010, CO013, CO015, CO016, CO017, CO028]关键公开可支撑指标显示其 Series D 前融资体量、临床阶段状态,以及公司声称的大数据足迹。员工数和收入仍未披露。
[CO010, CO012, CO013, CO015, CO019, CO022]1.4 平台架构与管线
PathOS 平台围绕三个引擎组织。Scout 是 AI 赋能的资产筛选引擎,扫描所有在研疗法,将每项疗法匹配到最可能受益的患者亚群,并输出排序列表。Sprint 指小型自治临床执行团队(Sprint Pods),在 Pathos 内部像独立迷你生物技术公司一样运作,专注把特定资产从一个里程碑推到下一个里程碑。Foundry 是共享 AI 核心,由与 Tempus 和 AstraZeneca 合作构建的肿瘤基础模型驱动;Foundry 也连接 Pathos 的实验室闭环 验证系统。 截至本次报告生成日,Pathos 管线包括四个项目。Pocenbrodib(P-300,原 FT-7051)是从 Novo Nordisk 于 2023 年引进的 CBP/P300 抑制剂(原开发方:Forma Therapeutics)。该项目于 2025 年 3 月 20 日进入转移性去势抵抗性前列腺癌 1b/2a 期临床试验(NCT06785636),目标入组约 203 名患者。P-500(PRT811)是 2024 年 8 月从 Prelude Therapeutics 引进的可入脑 PRMT5 抑制剂;1 期已完成,并在 IDH+ 高级别胶质瘤中显示经确认的完全缓解,Pathos 计划推进由生物标志物驱动的 2 期试验。DO-2 是来自 DeuterOncology(Belgium)的氘代第三代 MET 激酶抑制剂,Foundry 识别后,Pathos 于 2026 年 5 月取得多数股权并完成收购;1 期在可评估的 MET exon 14 跳跃突变 NSCLC 患者中显示 100% 肿瘤缩小,外周水肿为 5%,竞争药物为 62–82%。Milademetan(MDM2 抑制剂,来自 2024 年 1 月 Rain Oncology 收购)曾处 3 期阶段,但收购后未再出现在 Pathos 公开沟通中,引出信息新鲜度和管线重心问题。 Pathos 声称可访问超过 200 PB 多模态肿瘤数据,约为 The Cancer Genome Atlas(TCGA)规模的 50 倍。这些说法来自公司口径,尚未获得独立验证。[CO018, CO019, CO020, CO021, CO022, CO023]
Pathos AI 的商业模式把自研 AI 核心与外部数据合作、临床执行团队和逐步扩大的管线串起来,并全部回流到 Foundry 学习飞轮。
[CO018, CO019, CO021, CO024, CO028, CO038]1.5 时间线、里程碑与不利背景
Pathos AI 从 2022 年创立到 2026 年 5 月收购 DeuterOncology,不到四年内经历四次融资、四项资产新增、一次完成收购和两项重大战略合作。相对其公司规模,这一临床与企业发展节奏很快。里程碑表汇总了完整公开记录。 Rain Oncology 收购(2024 年 1 月 26 日完成)是首项重大外部交易。Pathos 全资子公司 WK Merger Sub, Inc. 以每股 $1.16 加每股或有价值权收购 Rain 全部流通普通股;股东投出 28,031,182 股,占流通股约 77%。Rain 的 milademetan 是 p53-MDM2 复合物抑制剂,曾推进至 3 期,但收购后基本从 Pathos 公开沟通中消失,可能意味着管线重新排序,也可能是战略性搁置。 2025 年 4 月 23 日宣布的 Tempus/AstraZeneca 合作,是单项资本消耗最大的事件:$200M 费用流向 Tempus,产出的基础模型将在三方之间共享。AstraZeneca 同时作为客户和模型共同开发方直接参与,形成少见的利益一致性,也需要开展治理和利益冲突尽调。The ASCO Post 强调过,在肿瘤决策场景部署 AI 工具存在法律与责任不确定性;这一点广泛适用于 Pathos 的目标。 三个风险点需要强调:第一,对 Huerga 的关键人依赖较高(新任 CEO,过往没有 Pathos 特定任期);第二,Series D 投资者身份未披露;第三,AI 驱动临床试验设计的监管和责任框架在全球尚未定型。[CO007, CO025, CO026, CO027, CO028, CO029]
| 日期 | 事件 | 类型 | 金额 / 估值 / 状态 | 参与方 | 含义 |
|---|---|---|---|---|---|
| 2022 | 由 Eric Lefkofsky 和 Ryan Fukushima 创立 | 创立 | 在特拉华州注册成立 | Eric Lefkofsky、Ryan Fukushima | 建立 AI 驱动的肿瘤药物开发公司;借力 Tempus 数据专长 |
| 2023-03 | 向 SEC 提交首份 Form D(相当于 Series A/B) | 融资 | $40M 融资(Series C 前累计) | Lightbank、Builders VC 等 | 资助 PathOS 平台初建和首批资产搜寻 |
| 2023 | 从 Novo Nordisk 获得 FT-7051(P-300/pocenbrodib)全球授权 | 产品 | 授权条款未披露 | Pathos AI、Novo Nordisk | 管线首个临床阶段资产;针对前列腺癌的 CBP/P300 抑制剂 |
| 2023-12 | 宣布以 $1.16/share + CVR 对 Rain Oncology(Nasdaq: RAIN)发起要约收购 | 产品 | $1.16/share + 或有价值权 | Pathos AI(通过 WK Merger Sub)、Rain 股东 | 启动 milademetan(MDM2 抑制剂,3 期)收购 |
| 2024-01 | Rain Oncology 收购完成;Rain 从 Nasdaq 退市 | 产品 | 28.03M 股(~77%)接受要约;Rain 成为全资子公司 | Pathos AI、Rain Oncology 股东 | Milademetan 纳入管线;Pathos 平台下首个 M&A 执行 |
| 2024-08 | 从 Prelude Therapeutics 获得 P-500(PRT811)全球授权 | 产品 | 授权条款未披露 | Pathos AI、Prelude Therapeutics | 新增可进入 2 期的脑渗透 PRMT5 抑制剂,面向胶质瘤和葡萄膜黑色素瘤 |
| 2024-10 | 完成 $62M Series C;NEA 领投,投后估值 $600M | 融资 | $62M,投后估值 $600M | NEA(领投)、Revolution Growth、Lightbank、Builders VC | 累计融资达 ~$102M;加速团队和平台扩张 |
| 2025-03 | pocenbrodib 1b/2a 期试验(NCT06785636)首例患者给药 | 监管 | 启动 203 名患者试验 | Pathos AI 临床团队、mCRPC 患者 | 核心项目进入活跃临床测试;精准生物标志物策略启动 |
| 2025-04 | 宣布与 AstraZeneca 和 Tempus AI 达成多年战略合作 | 伙伴关系 | 向 Tempus 支付 $200M 数据许可和模型开发费用 | Pathos AI、AstraZeneca (AZN)、Tempus AI (TEM) 三方合作 | 基础模型开发启动;业内最大肿瘤多模态模型项目 |
| 2025-05 | 完成 $365M Series D;Iker Huerga 出任 CEO | 融资 | $365M,投后估值 ~$1.6B | 新老投资者组合(未披露) | 最大一轮融资;管理层交接给有 AstraZeneca 背景的肿瘤老将 |
| 2026-05 | 收购 DeuterOncology 多数股权;DO-2(MET 抑制剂)纳入管线 | 产品 | 多数股权;条款未披露 | Pathos AI、DeuterOncology(比利时) | DO-2 由 Foundry AI 平台识别;首个完全通过 AI 执行的管线决策 |
事件日期和金额来自 Pathos 新闻稿、BioSpace、FierceBiotech 和 SEC Form D 文件。Series C 前 $40M 融资是根据 Series C 新闻稿推断,该稿称 $62M Series C 后累计融资达 $102M。里程碑类型使用报告 schema 允许的词表。
[CO004, CO005, CO010, CO011, CO013, CO015]Pathos AI 在不到四年里从创立推进到临床阶段公司,拥有四项资产、两项合作,累计融资 $467M;公司借助 AI 引导的资产筛选,压缩了传统开发时间线。
[CO004, CO007, CO010, CO011, CO013, CO021]1.6 展项
02市场分析
2.1 市场边界与范围
Pathos AI 位于三个分析上可拆分、战略上彼此咬合的肿瘤市场交汇处。第一,AI 辅助肿瘤药物发现:买方是大型和中型制药研发机构,目标是更高效识别并验证临床资产;现有替代方案包括内部生物信息学团队、与癌症中心的学术合作,以及传统统计方法做候选筛选。第二,肿瘤真实世界证据(RWE):生物医药公司、卫生技术评估机构和监管方购买回顾性或前瞻性患者数据集,用于支持监管申报、上市后监测和支付方谈判。第三,精准肿瘤诊断:二代测序(NGS)、多组学画像和伴随诊断既是临床照护层,也为 AI 训练提供数据底座。Pathos AI 直接可服务市场不包括药品制造、临床 CRO 运营服务、医院 EHR 软件和药品销售收入——这些都相邻,但不属于平台许可模式。癌症流行病学背景很大:NCI 估计美国 2025 年新增癌症诊断约 2.04M 例,WHO 报告 2022 年全球约 10M 例癌症死亡,SEER 预计 2026 年美国新增约 2.11M 例和 626K 例癌症死亡。这样的患者规模创造了纵向数据机会,肿瘤 AI 平台竞相聚合。市场边界高度有争议:Tempus AI 和 Caris Life Sciences 都同时覆盖诊断、RWE 和 AI 建模,因此竞争市场划分取决于假设;若不调整口径,不同分析师估计不可直接比较。[CM007, CM008, CM009, CM021, CM022]
| 细分 / 类别 | 纳入支出 | 排除支出 | 买方 / 支付方 | 与 Pathos AI 的相关性 |
|---|---|---|---|---|
| AI 辅助肿瘤药物发现 | 平台许可费、数据访问订阅、AI 模型训练算力 | 药品生产、临床 CRO 服务、药品销售收入 | 大型制药公司研发(CDO / VP R&D) | 核心市场 — PathOS Scout 和 Sprint 直接瞄准该细分 |
| 肿瘤真实世界证据(RWE) | RWE 数据库订阅、分析服务、回顾性队列访问 | EMR 系统软件、医院 IT 基础设施 | 制药 / 生物科技(医学事务、HEOR)、监管机构 | 紧邻市场 — Pathos 声称拥有 200+ PB 多模态数据 |
| 精准肿瘤诊断 | NGS 分析、伴随诊断(CDx)、液体活检、IHC 面板 | 药品报销、医院 COGS | 医院、实验室、支付方(报销) | 数据底座 — 生成支撑 AI 训练的多模态数据 |
| 临床试验优化(AI) | 试验设计工具、患者匹配 AI、去中心化试验平台 | CRO 运营执行成本 | 生物科技 / 制药临床运营 | 邻近市场 — PathOS Sprint 参与试验设计 AI 竞争 |
| 肿瘤治疗药物(药品收入) | 已获批肿瘤药物收入 | (以上类别) | 医院处方集、支付方、患者 | 范围外 — Pathos 尚无药品收入;管线仍处于获批前 |
市场边界存在争议;Tempus AI 和 Caris Life Sciences 都同时跨越多个细分。纳入 / 排除支出分类是用于界定范围的分析构造,不是合同定义。截至 2025 年 5 月 Series D 交割,Pathos AI 没有披露产品收入。
[CM007, CM008, CM009, CM021]2.2 市场规模测算:多重视角
测算 Pathos AI 可触达机会,需要拆开多个彼此重叠的发布方市场估计。IQVIA《Global Oncology Trends 2025》报告 2024 年全球肿瘤药物支出为 $252B,预计到 2029 年达到 $441B,CAGR 约 11.9%;这是 AI 平台服务争夺一部分研发预算的整体肿瘤经济。MarketsandMarkets 估计专门的肿瘤 AI 市场 2024 年为 $2.45B,到 2030 年增至 $11.52B,CAGR 29.4%,范围涵盖药物发现、诊断、临床决策支持和影像 AI。药物发现技术市场(AI 只是其中一个细分)单独估计为 2025 年 $30.58B,到 2030 年 $51.51B,CAGR 11.0%(MarketsandMarkets,2026 年 1 月)。Mordor Intelligence 估计 2025 年全球 RWE 解决方案市场为 $2.44B,其中肿瘤占行业 34.65%;由此推算 2025 年肿瘤 RWE 细分约 $846M,并以 16.33% CAGR 增长,到 2031 年总市场达到 $6.04B。精准诊断与医学市场在 2024 年达到 $145.53B(MarketsandMarkets),NGS 市场 2023 年为 $12.98B,并以 18.3% CAGR 增长(Allied Market Research)——两者都说明 AI 药物发现依赖的数据底座市场本身在快速扩张。这些估计不确定性很高:发布方采用不同边界定义,没有分析师独立验证竞争对手方法论,肿瘤 AI 子细分边界尤其多孔。IQVIA 还报告,PD-1/PD-L1 骨干疗法开始面对生物类似药竞争后,肿瘤支出增长预计在 2027 年后放缓,为 2028–2029 年总可用市场增长增加结构性逆风。分析师市场规模估计相差超过 50 倍,从肿瘤 RWE 切片(约 $846M)到完整肿瘤药物经济($252B),展示了市场边界定义必须解决的范围假设跨度。[CM001, CM002, CM003, CM004, CM005, CM006]
| 发布方 | 年份 | 地理范围 | 市场 | 数值(基准年) | 预测值 | CAGR | 置信度 | 关键限制 |
|---|---|---|---|---|---|---|---|---|
| IQVIA | 2025 | 全球 | 肿瘤药物支出 | $252B (2024) | $441B (2029) | ~11.9% | 中 | 仅药品销售审计;不含 AI 平台费和数据服务 |
| MarketsandMarkets | Dec 2024 | 全球 | 肿瘤领域 AI | $2.45B (2024) | $11.52B (2030) | 29.4% | 低 | 边界包括影像 AI 和诊断 AI,也包括药物发现 |
| MarketsandMarkets | Jan 2026 | 全球 | 药物发现技术 | $30.58B (2025) | $51.51B (2030) | 11.0% | 低 | 包括仪器、试剂、软件;AI 只是其中未量化的子集 |
| Mordor Intelligence | 2026 | 全球 | RWE 解决方案(肿瘤部分:34.65%) | $2.44B 总额;肿瘤约 $846M (2025) | $6.04B 总额 (2031) | 16.33% | 低 | 肿瘤占比由分析师套用细分比例得出;没有子细分审计 |
| Allied Market Research(市场研究机构) | 2023 | 全球 | NGS 市场 | $12.98B (2023) | $97.81B (2035) | 18.3% | 低 | 覆盖所有 NGS 应用;肿瘤子集未定义 |
| MarketsandMarkets | Jan 2025 | 全球 | 精准诊断和精准医疗 | $145.53B (2024) | $246.66B (2029) | 11.1% | 低 | 范围很宽,包括治疗和非肿瘤诊断 |
| GrandView Research | 2026 | 全球 | 临床试验中的 AI | 可访问内容未披露 | 预计显著增长 | 未披露 | 低 | 抓取页面没有给出具体数值;仅作方向性估计 |
各项估计只有低到中等置信度,因为不同发布方对市场边界的定义不一致。没有两家发布方使用可比的方法或边界。肿瘤 RWE 数字为推导值(34.65% × $2.44B)。CAGR 不能跨行相加;已发布 CAGR 数字按原文采用。
[CM001, CM002, CM003, CM004, CM005, CM006]Pathos AI 可服务肿瘤 AI 市场的三层 TAM/SAM/SOM 金字塔。TAM 锚定肿瘤学 AI(MarketsandMarkets 2024)。SAM 限定在面向生物制药的 AI 药物发现服务(估算子集)。SOM 是对 Pathos AI 近期可获取份额的数量级估算,处于商业化前。
TAM 取 MarketsandMarkets 2024 肿瘤学 AI($2.45B),四舍五入为 $2,450M。SAM 是分析性估算(约为 AI 肿瘤学 TAM 的 16-20%),代表仅用于药物发现的生物制药平台支出,排除影像 AI、诊断 AI 和临床决策支持。~$50M 的 SOM 是 Pathos AI 当前阶段的数量级估算,类比早期商业化 AI 药物发现平台;Pathos 在治疗药物上仍处于商业化前,也没有披露平台收入。
[CM001, CM002, CM003, CM036]围绕 Pathos AI 的四个市场规模视角给出低 / 基准 / 高估算,均按各自预测期折算为十亿美元。图表展示分析师不确定性和边界敏感性。
肿瘤药物支出区间锚定 IQVIA 中点,并用 ±10% 边界反映 CAGR 不确定性和生物类似药逆风情景。AI 肿瘤学区间锚定 MarketsandMarkets 中点;考虑边界不确定性,采用 ±25% 边界。药物发现区间取 MarketsandMarkets 中点 ±15%。肿瘤学 RWE 区间来自 Mordor 总市场预测,并套用 34.65% 肿瘤份额,再按 ±50% 份额不确定性处理。所有数字都是方向性估算;无法对不同发布方的方法论做交叉校准。
[CM001, CM002, CM004, CM005, CM037]2.3 买方与细分市场格局
肿瘤 AI 数据和药物发现平台的主要买方是大型制药公司,由首席数字官和 VP 级研发领导分配多年期数据许可和平台费用。商业规模证据较强:Tempus AI 报告 2026 年 Q1 收入 $348.1M(同比 +36.1%),全年 2026 指引为 $1.59–1.60B,并与 Merck、Gilead、AstraZeneca 等开展多年战略合作,为企业提供肿瘤 AI 数据访问。Caris Life Sciences 报告 2026 年 Q1 收入 $216.2M(同比 +79%),完成 52,800 例临床分型,数据库累计画像超过 1.07M,显示有大型临床装机基础为分子分型服务付费。Tempus 与 Caris 合计指向约 $2.3B 年化肿瘤数据服务市场,锚定现实商业规模。ConcertAI 是另一家肿瘤 RWE 平台,2024 年 6 月以约 $1.9B 估值融资 $150M;Roche 2018 年以约 $1.9B 收购 Flatiron Health,较早证明生物医药愿意为肿瘤 RWD 平台支付高溢价。中型肿瘤生物技术公司是第二层买方,用 AI 提高资产筛选和试验设计效率,从而延展有限资本。学术癌症中心通过赠款资助协议大量购买数据平台,但很少产生商业规模合同。社区肿瘤网络和支付方(保险)正在成为 AI 导航服务买方——Humana 于 2025 年与 Thyme Care 合作——但仍处在 Pathos AI 当前销售动作外围。Pathos AI 声称拥有 200+ PB 多模态数据(约 50× TCGA),但该数字似乎只出现在公司材料中,没有独立验证,因此数据集差异化主张难以评估。制药研发预算分配是关键门槛;已经验证的模式是多年期企业协议,预算线在研发(非 COGS),供应商选择由首席数字或 AI 负责人控制。[CM013, CM014, CM015, CM016, CM017, CM018]
| 细分市场 | 买方 | 用户 | 付款方 | 工作流 | 预算负责人 | 采用触发因素 |
|---|---|---|---|---|---|---|
| 大型生物制药(Top-20 制药公司) | Pfizer、J&J、AstraZeneca、Merck、Roche、Gilead 等药企 | 药物发现科学家、数据科学团队、CDO 办公室 | R&D 预算(通常每年为 $B 量级) | 管线筛选、生物标志物识别、试验设计 | 首席数字官 / 研发高级副总裁 | AI 竞争压力、管线失败率、基础模型联合开发 |
| 中型肿瘤生物技术公司 | 专注肿瘤、上市市值 $100M-$2B 的生物技术公司 | 临床和转化研究团队 | 风险投资、合作收入 | 资产选择、试验优化、数据访问 | CEO / CMO | 需要提高资金效率;高淘汰成本形成压力 |
| 学术癌症中心 | NCI 指定癌症中心、大学医院 | 教职研究员、博士后、临床研究者 | 资助资金(NIH、NCI、基金会) | 患者队列研究、回顾性研究、模型验证 | 首席研究员 / 系主任 | 数据访问、研究产出、资助交付物 |
| 社区肿瘤网络 | US Oncology Network、American Oncology Network 等肿瘤网络 | 肿瘤医生、病理医生 | 保险报销(CMS、商业保险) | 用于治疗选择的分子分型、依从性跟踪 | 医疗总监 / CMO | CMS 对 CGP 检测的报销、质量改进计划 |
| 支付方和护理导航公司 | Humana、Cigna、UnitedHealth 等支付方;Thyme Care、Included Health 等护理平台 | 临床运营和分析团队 | 保费收入、风险池 | AI 导航、肿瘤权益管理、成本控制 | 临床创新副总裁 | 价值医疗合同、监管报告要求 |
买方 / 付款方角色分配来自已披露合作和可比肿瘤 AI 公司的商业模式类比。Pathos AI 未披露预算分配数字。学术和社区买方的病例量更大,但单个合同支出低于制药企业级交易。
[CM013, CM014, CM015, CM017, CM018, CM022]肿瘤 AI 数据市场中,买方细分(行)与商业参与维度(列)的矩阵。取值反映各细分对 Pathos AI 平台产品的定性参与强度。
[CM021, CM022, CM023, CM024, CM038]2.4 增长驱动因素与采用约束
多股结构性力量推动 AI 赋能肿瘤药物发现需求。药物开发失败率仍极高——肿瘤药从 1 期到监管获批的成功率约 5%——这给用 AI 改进资产筛选和试验设计创造了强商业动机(Springer 2025 综述)。IQVIA 报告 2024 年肿瘤试验启动 2,162 项(较 2019 年 +12%),2024 年全球推出 25 种新型肿瘤活性物质,2020–2024 年平均每年 26 种,高于 2015–2019 年每年 16 种,说明管线生产力持续,AI 优化工具仍有需求。FDA 的 Real-World Evidence Framework(2018)及后续指南正式为部分申报打开 RWE 证据标准,形成监管顺风。NGS 测序成本下降,让大规模多组学数据集以更低单样本成本成为可能;NGS 市场预计从 2023 年 $12.98B 增至 2035 年 $97.81B,CAGR 18.3%。Tempus–AstraZeneca–Pathos AI 合作构建最大多模态肿瘤基础模型(2025 年 4 月,向 Tempus 合计支付 $200M),体现大型药企正在主动共同投资 AI 数据基础设施。GrandView Research 预计 AI 在临床试验中的增长显著,驱动因素包括更快招募、合成对照臂和生物标志物入组标准需求。采用约束同样实质且持久。HIPAA(美国)和 GDPR(欧盟)限制跨机构数据流动;联邦学习等隐私保护架构(Owkin 已采用)正在成为商业前提。既有平台 Tempus AI 与 Caris Life Sciences 已建立稳固制药关系,切换成本高:多年期纵向研究项目把平台数据嵌入多个治疗项目,迁移昂贵且缓慢。AI 生成药物发现假设尚无明确 FDA 监管路径,平台衍生候选药物面临审批路径不确定性。自研多模态数据集搭建资本强度高——Pathos AI 未披露数据集成本——而 Pathos 声称拥有 200+ PB 数据集缺乏独立验证,构成尽调缺口。Springer(2025)综述还指出,肿瘤 AI 主要在单中心数据上做回顾性验证,限制真实世界泛化能力。[CM020, CM025, CM026, CM027, CM028, CM029]
| 驱动因素 / 约束 | 方向 | 时间 | 对 Pathos AI 的影响 | 尽调追问 |
|---|---|---|---|---|
| 肿瘤药物失败率高(从 Phase 1 到获批的淘汰率约 95%) | 驱动 | 当前 | 催生 AI 辅助资产选择和试验优化的迫切商业需求 | Pathos AI 是否验证过,Scout 选择的资产淘汰率优于行业基线? |
| FDA RWE Framework 和持续接受 RWE 指引 | 驱动 | 当前 / 近期 | 支持数据驱动型监管申报;扩大 RWE 解决方案市场 | Pathos 数据是否参与过任何监管申报?如有,是哪些? |
| NGS 测序成本下降,支持大规模多组学数据集 | 驱动 | 当前 / 长期 | 降低单样本数据获取成本;支撑更大的训练数据集 | Pathos AI 的单样本数据获取成本是多少,趋势如何? |
| 精准医疗要求和伴随诊断增长 | 驱动 | 当前 | 推动临床对多组学分型的需求,而多组学分型支撑 AI 模型训练 | 多少个 FDA 批准的伴随诊断引用了 PathOS 中的数据? |
| HIPAA 和 GDPR 数据隐私法规 | 约束 | 当前 / 持续 | 限制跨机构数据流动;要求隐私保护型架构 | PathOS 使用什么数据治理和去标识化架构? |
| 现有厂商切换成本(Tempus、Caris 嵌入纵向数据) | 约束 | 中期 | 制药合作伙伴把平台数据嵌入多年研究项目;替换成本高 | Tempus/Caris 与制药合作伙伴使用哪些合同锁定机制? |
| AI 生成药物发现假设的 FDA 路径尚未定义 | 约束 | 近期 / 演进中 | 平台衍生药物候选物的获批路径存在不确定性 | 是否有 Pathos 药物候选物受到 FDA 针对 AI 的监管审查? |
| 多模态数据组装和计算的资本强度 | 约束 | 当前 | 烧钱速度高;数据集成本未披露;Series D 资金有限 | Series D 后,Pathos AI 的季度烧钱速度和现金跑道是多少? |
时间判断是定性估计。驱动因素 / 约束评估基于行业类比和已披露合作,不是 Pathos AI 运营数据披露。尽调追问仍是开放问题,需要管理层沟通或独立技术审查。
[CM020, CM025, CM026, CM027, CM028, CM029]肿瘤 AI 数据平台的示意性生物制药采用漏斗,以 Tempus AI 已观察到的商业轨迹作为市场参照。每一阶段代表一个独立参与关口,并估算制药 R&D 组织的可触达基数。
漏斗数值是数量级估算,基于 Tempus AI 已披露制药合作伙伴数量和行业类比;不是 Pathos AI 专属数据。AstraZeneca、Pathos AI 和 Tempus AI 是漏斗底部已观察到的基础模型共同开发伙伴。其他阶段均为示意。
[CM025, CM030, CM031, CM039]03竞争格局
3.1 格局结构:Pathos 夹在发现引擎与肿瘤数据既有玩家之间
Pathos 周围的竞争图谱比普通生物技术同业更宽,因为 Pathos 想把几类工作压进一个运营模型。Pathos 公开材料将公司表述为重新定义肿瘤药物开发、推进精准构建管线,而不是销售独立软件席位,也不是从零发明每个分子。这意味着直接可比对象不只是其他肿瘤生物技术公司。Pathos 与 Recursion/Exscientia、Insilico、Schrödinger、Relay 等 AI 优先发现平台竞争技术可信度、资本和差异化资产。同时,它也与 Tempus、Flatiron、Caris、ConcertAI、Foundation Medicine 等肿瘤数据和工作流既有玩家争夺多模态数据、真实世界证据和临床医生工作流分发。Owkin、Valo、Boundless、Absci 以及原 Ikena 平台等相邻 AI 生物技术玩家也重要,因为它们显示了行业走向,以及资本、数据或证明点变化时竞争压力会多快转向。结果是一个夹心位置:Pathos 比多数发现同业更贴近临床,但商业化程度远低于它也必须学习或围绕合作的肿瘤数据既有玩家。[CP001, CP002, CP027, CP028, CP029, CP030]
| 竞争对手 | 类别 | 规模 / 状态 | 目标细分市场 | 差异化 | 局限 |
|---|---|---|---|---|---|
| Pathos AI | AI 辅助肿瘤资产开发 | 私营、临床阶段管线搭建者 | 生物标志物定义的肿瘤开发 | 在肿瘤领域内部组合资产分诊、生物标志物逻辑和执行 | 没有公开完成的闭环成功案例,也没有独立软件收入 |
| Recursion + Exscientia | AI 优先的端到端药物发现 | 合并后的上市 AI 药物发现平台 | 广泛发现合作和内部管线 | 表型组学和生成式化学放在同一屋檐下 | 整合和资本消耗风险仍实质存在 |
| Insilico Medicine | AI 原生靶点到分子平台 | 40+ 项目;13 条 IND 获批管线;Lilly 验证 | 多治疗领域发现和授权 | 端到端 Pharma.AI 技术栈,授权信号强 | 临床证据仍集中在早期项目 |
| BenevolentAI | 知识图谱 TechBio / 合作发现 | 2024 年重置后走向重组和退市 | 合作发现加规模更小的内部组合 | 生物医学知识图谱和靶点假设引擎 | 战略收缩削弱了竞争威胁 |
| Schrödinger | 基于物理的软件与管线 | 上市软件和治疗混合体 | 制药计算设计和内部肿瘤项目 | 经常性软件授权加自有项目 | 对试验执行或资产修复的聚焦较弱 |
| Relay Therapeutics | 基于运动的精准肿瘤学 | 上市临床阶段平台,选择性对外授权 | 肿瘤和罕见病项目 | 用于基于运动药物设计的 Dynamo 平台 | 比数据平台护城河更窄 |
| Tempus AI | 临床基因组数据和肿瘤工作流 | 上市 AI 医疗科技平台 | 检测、诊断、制药数据、肿瘤诊所 | 通过分型和 EMR 集成铺开工作流 | 不是肿瘤药物资产的所有者兼运营者 |
| Foundation Medicine | CGP 诊断平台 | Roche 关联公司 | 伴随诊断和基因组分型 | 覆盖 300 多个基因的 CGP 足迹和 Roche 触达 | 比部分数据平台更偏诊断中心,也更少多模态 |
| ConcertAI | 肿瘤数据和企业 AI | 私营企业平台 | 生命科学研究和商业团队 | 基于肿瘤数据集的 CARAai 和 Precision Suite | 更像软件 / 数据卖方,而不是治疗开发者 |
| Flatiron Health | 肿瘤工作流和 RWE 引擎 | 大型肿瘤工作流现有厂商 | 肿瘤诊疗机构和研究用户 | OncoEMR 分发加研究规模 | 自身不拥有治疗资产 |
| Caris Life Sciences | 分子分型和多模态研究数据 | 大型精准肿瘤平台 | 诊断和生物制药研究 | DNA、RNA 和影像结合合作伙伴工作流 | 诊断优先姿态限制了与 Pathos 的直接相似度 |
| Owkin | 端到端 AI 生物技术和诊断 | 私营 AI 生物技术公司,与制药企业合作 | 精准医疗、诊断和发现 | 因果 AI 加病理和空间组学合作 | 疾病范围比 Pathos 当前切入点更宽 |
| Valo Health | AI 驱动的因果生物学公司 | 私营平台,领导层面向制药 | 广泛疾病领域和合作发现 | 基于大规模人类数据的闭环化学 | 肿瘤专属性弱于 Pathos |
| Boundless Bio | 特定生物学机制的肿瘤替代方 | 临床阶段精准肿瘤公司 | ecDNA 扩增癌症 | 独特的 ecDNA 生物学切入点 | 单一生物学焦点缩窄可触达市场 |
| Ikena / ImageneBio | 曾经相邻的肿瘤平台 | 合并后 Ikena 不再独立 | 合并后聚焦 anti-OX40 和免疫学 | 体现资本回收并注入新实体 | Ikena 不再作为独立实体威胁 Pathos |
| Absci | 相邻的 AI 设计生物制剂平台 | AI 生物制剂管线,含 Phase 1/2a 资产 | 生物制剂和炎症,不以肿瘤为先 | 展示 AI 原生生物制剂设计的相邻性 | 位于 Pathos 当前肿瘤小分子赛道之外 |
覆盖范围有意保持部分、偏战略而非穷尽;该表强调的是与 Pathos 2026 年公开定位最相关的同行和替代方。
[CP003, CP004, CP006, CP007, CP010, CP012]这张图用顺序评分衡量 AI 原生发现深度,以及临床和商业证据;Pathos 处在发现优先的 AI 生物科技公司与已规模化肿瘤数据平台之间。
[CP003, CP006, CP009, CP014, CP017, CP020]3.2 直接 AI 药物开发同业:发现深度更强,临床证明不一,耐久性参差
在直接 AI 药物开发同业中,没有一家与 Pathos 完全匹配,但每一家都压迫 Pathos 论点的不同部分。Recursion 加 Exscientia 是最清晰的端到端发现重量级玩家:两家公司合并时明确主打具备端到端能力的合并领导者,因此成为 AI 规模和生成式化学同处一屋檐下的最佳公开案例。Insilico 在纸面上拥有最清晰的 AI 平台广度,项目超过 40 个、13 条已获 IND 放行的管线,并与 Lilly 达成规模足够像真实生物医药验证事件的合作。Schrödinger 又不同:它是经济模型最容易看懂的同业,因为软件许可和合作经济收益可见,商业模式看起来比多数 AI 生物技术故事更耐久。Relay 的强项不是泛化 AI 叙事,而是基于运动的蛋白科学和正在临床推进的肿瘤管线,包括已对外授权资产。BenevolentAI 是警示案例。其 Merck 合作显示真实伙伴验证,但 2024 年战略重置和 2025 年退市路径说明,一旦资本市场和管线进展无法互相强化,AI 生物技术 叙事会迅速收缩。与这一组相比,Pathos 在临床资产分诊和生物标志物指导开发上更有辨识度,但分子设计深度尚未充分证明。[CP003, CP004, CP005, CP006, CP007, CP008]
| 采购标准 | Pathos | Recursion / Exscientia | Insilico | Schrödinger | Relay | Tempus / Flatiron |
|---|---|---|---|---|---|---|
| 新靶点识别 | 中等 | 强 | 强 | 中等 | 中等 | 弱 |
| 分子设计所有权 | 弱 | 强 | 强 | 强 | 强 | 弱 |
| 临床资产分诊 / 修复 | 强 | 中等 | 弱 | 弱 | 弱 | 弱 |
| 试验设计 / 患者选择 AI | 强 | 初现 | 中等 | 弱 | 弱 | 中等 |
| 肿瘤诊疗工作流 / 分发 | 初现 | 弱 | 弱 | 弱 | 弱 | 强 |
| 商业软件 / 数据变现 | 弱 | 弱 | 初现 | 强 | 弱 | 强 |
单元格是基于证据的序位判断,综合了官方平台页面、合作伙伴公告和公司披露,而不是来自供应商比较页面。
[CP005, CP006, CP008, CP009, CP014, CP016]Pathos 在资产分诊和试验设计 AI 上最强;同行通常在分子设计、软件变现或数据工作流分发上更强。
[CP005, CP008, CP014, CP016, CP018, CP027]3.3 数据平台与相邻替代品:最难察觉的竞争来自分发与证据
Pathos 面临的非显性威胁不只是另一家 AI 发现初创公司,而是能够决定哪些资产获得证据、采用和制药关注的肿瘤数据及工作流层。Tempus 已经嵌入分子分型,并证明它能直接接入 Flatiron 的 OncoEMR;Caris 和 Flatiron 也已把基因组、影像和真实世界数据组合成商业研究产品。ConcertAI 从企业侧沿同一模式推进,为生命科学用户打包肿瘤数据集和智能体 AI 工具,而不是面向单一管线投资者。Foundation Medicine 模态更窄,但仍然强大,因为 Roche 将其定位在 300 多个癌症驱动基因和全球 CGP 采用之上。相邻 AI 生物技术玩家进一步扩大威胁面。Owkin 结合因果 AI、病理、空间组学和制药合作;Valo 推进 AI 赋能的人类因果生物学和闭环化学;Boundless 聚焦 ecDNA 肿瘤切入点;Absci 展示 AI 设计生物制剂可能是什么样;Ikena 并入 ImageneBio 则提醒,资本再配置能在一个表面同业成为耐久竞争者前就将其抹去。对 Pathos 来说,教训是数据访问、工作流分发和伙伴信任可能与模型质量同样重要。[CP020, CP021, CP022, CP023, CP024, CP025]
| 公司或类别 | 观察到的合同模式 | 公开透明度 | 包含能力 | 含义 |
|---|---|---|---|---|
| Pathos AI | 无公开独立定价;价值通过自有和合作肿瘤项目积累 | 低 | 资产选择、生物标志物策略和开发执行 | 买方或投资人很难为外部变现设基准 |
| Recursion / Exscientia | 发现合作和里程碑驱动经济条款 | 中 | AI 发现平台,加合作和内部管线工作 | 制药企业兴趣提供验证,但仍以交易为主,而非按工作流定价 |
| Insilico Medicine | 首付款、里程碑、特许权使用费和组合授权 | 中 | 靶点发现、分子设计和项目对外授权 | 外部验证强,但没有简单的 SaaS 可比对象 |
| Schrödinger | 多年期软件许可,加里程碑付款和特许权使用费 | 较高 | 计算设计软件和协作发现 | 同业中最透明的经常性软件经济性 |
| Relay Therapeutics | 管线所有权,加选择性对外授权 | 中 | 平台发现和自有肿瘤资产 | 资产变现可反哺平台工作,但不形成广泛工作流定价 |
| Tempus / Flatiron | 检测、工作流和企业平台合同 | 中 | 分子分型、EMR 集成和研究工具 | 比 Pathos 更接近采购预算和运营锁定 |
| ConcertAI / Caris | 企业级肿瘤数据和 AI 研究协议 | 中 | 临床和基因组数据库,配套试验与证据支持 | 卖进生命科学预算,而不是讲风险投资式管线故事 |
| Owkin / Valo | 合作驱动的 AI 生物技术协作 | 低至中 | 精准医疗工具、发现合作和内部项目 | 最接近 Pathos 后续形态的相邻变现参照 |
这一领域公开标价稀缺;因此,包装方式只能通过披露的合作结构、集成和企业销售动作来比较,而不是按名义价卡比较。
[CP008, CP009, CP016, CP019, CP027, CP030]3.4 护城河耐久性:Pathos 有差异化赛道,但护城河仍更像可能性而非证明
支持 Pathos 的最强论点,是它试图占住一个狭窄但有价值、其他玩家没有真正拥有的层:购买或许可具备临床价值的肿瘤资产,用 AI 和多模态证据选择最佳患者和试验设计,再把产出的证据反馈到下一次组合决策。这是一条真实的战略赛道。问题在于,公开可见的切换成本和定价权仍然偏弱。Tempus、Flatiron、Caris 和 ConcertAI 更靠近临床工作流和生命科学预算。Schrödinger 在经常性软件变现上更清晰。Recursion、Insilico 和 Relay 在核心发现或设计上更深。因此 Pathos 需要展示一个市场看得见的闭环结果:由平台选择或重新设计的项目,明显优于传统资产管理可能带来的结果。在此公开出现前,护城河只能按新兴阶段假设。这个细分位点有战略吸引力,若被证明很可能具备耐久性,但耐久性尚未在已完成结果、重复合同或嵌入式工作流锁定中显现。[CP035, CP036, CP037, CP038, CP039, CP040]
| 护城河主张 | 威胁 | 严重性 | 现有证据 | 缓释措施 / 尽调要求 |
|---|---|---|---|---|
| 肿瘤学闭环学习 | Pathos 尚未公开展示过一个已跑通、并能改进下一次决策的飞轮 | 高 | 公开证据更多证明资产组装和合作公告,而不是结果闭环 | 索取项目层面的前后对照证据,覆盖生物标志物选择、试验设计和周期时间 |
| 独家数据优势 | Tempus、Flatiron、Caris、ConcertAI 和 Foundation 已经大规模控制数据或工作流入口 | 高 | 现有数据层今天已商业化,Pathos 还在证明复用权利和杠杆 | 审查数据权利排他性、合作伙伴经济条款和复用许可 |
| 资产筛选速度 | 发现优先的同业可以内部寻找或设计分子,而不是修复外部资产 | 中 | Recursion、Insilico、Schrödinger 和 Relay 都展示了更深的发现技术栈 | 将 Pathos 决策耗时和 BD 命中率与传统资产搜寻对标 |
| 肿瘤学专注 | 资本或生物学依据足够时,相邻 AI 生物技术公司也能切入肿瘤学 | 中 | Owkin、Valo、Absci 和 Boundless 展示了进入精准医疗的不同相邻技术路径 | 讲清楚 Pathos 超越一般 AI 生物技术叙事的独特优势在哪里 |
| 合作伙伴杠杆 | 依赖外部数据和平台伙伴,会削弱 Pathos 对关键输入的控制 | 高 | Pathos 比既有数据公司更依赖具名数据和模型构建关系 | 审查主要合作伙伴合同期限、排他性和终止触发条件 |
| 切换成本 | Pathos 尚未嵌入肿瘤诊所工作流或企业软件预算 | 高 | Tempus 和 Flatiron 已经更接近照护点 | 索取申办方、站点和合作伙伴复用数据,以及留存证据 |
| 资本市场耐久度 | 战略或临床受挫后,AI 生物技术同业可能迅速收缩 | 中 | BenevolentAI 和 Ikena 都说明,看似有威胁的玩家可以多快被重置或吸收 | 保守测算现金跑道和证据周期 |
这张登记表服务于承销判断,不做绝对排名;它突出的是,Pathos 声称的护城河要变得耐久,哪些条件必须成立。
[CP012, CP013, CP022, CP023, CP041, CP043]Pathos 今天看起来有竞争差异,但多数护城河指标仍只是早期成形,还谈不上守住。
[CP041, CP044, CP045, CP046, CP047, CP048]3.5 展项
04财务情况
4.1 融资历史与资本充足性
对 Pathos 做财务承保,要先看融资,而不是利润表质量,因为公司未公布财务报表或公开现金余额。不过,对一家私有生物技术公司来说,其资本记录锚点异常扎实。Pathos 于 2024 年 10 月宣布完成 $62M Series C,投后估值 $600M;2025 年 5 月完成 $365M Series D,投后估值约 $1.6B。CIK 0001967854 下的 SEC Form D 记录印证了绑定公司的三次豁免发行:2023 年文件显示发行总额约 $40M,2024 年文件匹配 Series C 规模,2025 年文件显示发行总额 $399,999,933,截至申报日已售出 $282,999,950,投资者 11 名。这组证据为核心融资时间线提供了异常强的一手锚点。 更重要的承保问题是未来资本充足性。Pathos 披露 Series D 资金将用于临床管线和肿瘤基础模型的持续投入,2025 年 4 月 Tempus/AstraZeneca 公告则单独披露应向 Tempus 支付 $200M 费用。已审阅的官方和监管文件来源中未出现债务、仓储融资或项目融资义务,因此资产负债表看起来由股权融资支撑。但缺少交割时现金余额,投资者无法判断 Series D 完成时此前资本还剩多少、Tempus 承诺付款速度多快,或已有多少资本承诺给资产收购和试验执行。因此,本章将融资历史视为已验证,将现金充足性视为估计。[CI001, CI002, CI003, CI004, CI005, CI006]
| 项目 | 日期 / 阶段 | 公开金额 / 数值 | 公开对手方 | 承销含义 | 来源质量 |
|---|---|---|---|---|---|
| 初始 Form D 发行 | 2023 年申报文件 | 总发行 $39,999,988;申报时已售 $19,999,992;4 名投资者 | 申报文件未公开列名投资者 | 确立 Series C 前约 $40M 资本基础 | 申报文件 |
| Series C 前推算资本基础 | 截至 2024 年 10 月 | Series C 前累计融资约 ~$40M | 根据 Series C 总融资声明和 2023 Form D 推导 | 确认公司进入 Series C 时资本基础相对有限 | 官方 + 申报文件 |
| Series C 融资 | 2024-10 / 2024-11 | $62M,投后估值 $600M | NEA 领投;Revolution Growth 和现有内部投资者也已披露 | 重大估值跃升,为主要试验启动前的管线扩张提供资金 | 官方 + 申报文件 |
| Series C Form D 记录 | 首次销售 2024-10-24 | 总发行 $61,999,979;13 名投资者 | 申报文件未披露投资者姓名 | 一手层面印证:已宣布融资转化为已申报发行 | 申报文件 |
| Series D 宣布交割 | 2025-05-15 | $365M,投后估值约 $1.6B | 新老投资者混合;名单大多未披露 | 融资规模很大,但缺少公开收入,估值仍需独立承销 | 官方 |
| Series D Form D 记录 | 提交于 2025-05-01 | 总发行 $399,999,933;已售 $282,999,950;11 名投资者 | 申报文件未披露投资者姓名 | 表明该轮可能在完全交割前已申报,且公告总额与申报已售金额存在时间差 | 申报文件 |
| Tempus 费用承诺 | 披露于 2025-04-23 | $200M | Pathos/AstraZeneca 合作下的 Tempus AI | 已知现金流出降低了清晰现金桥的可见度 | 官方 |
| 估计年度烧钱基准 | 2025-2026 情景 | 每年 ~$120M-$240M | 基于肿瘤学 / AI 生物技术公开基准 | 表明 Pathos 资金充足,但相对于其雄心并未明显过度资本化 | 估计 |
| Series D 后估计独立现金跑道 | 交割时情景 | ~18-36 个月 | 基于 Series D 规模和基准烧钱区间 | 有用的方向性框架,不是已验证现金跑道 | 估计 |
| 债务 / 项目融资义务 | 运行日视角 | 公开资料未披露 | N/A | 股权融资看似占主导,但没有披露不等于没有承诺 | 观察所得 |
融资事实证据很强;资本充足性结论并不强。没有交割现金余额和已披露义务付款时间表,现金跑道仍是情景分析,不是已验证数字。
[CI001, CI002, CI003, CI004, CI005, CI006]展示 Pathos 如何把股权融资转化为发现、开发和合作方义务,以及为何缺失的期末现金余额仍是承保的闸门变量。
节点描述公开来源可见的资本去向和义务。它们不是管理层资金分配,也不是经审计的现金用途表。
[CI009, CI010, CI011, CI012, CI033, CI034]4.2 收入模型、商业化动作与公开牵引力
Pathos 不应按具备可观察商业队列的软件公司建模。公开叙事是一家临床阶段肿瘤平台:收购或许可资产,用多模态数据和 AI 改进试验设计,最终通过合作、许可、里程碑、版税或资产退出捕获价值。这可能在经济上有吸引力,但截至本次报告生成日,本章审阅的公开来源没有披露 ARR、确认收入、收入结构、软件定价、服务定价、续约率、客户数量,或任何其他能支撑传统收入质量评估的当前变现指标。 唯一公开披露的具体合作经济收益方向相反:Pathos 称其 Tempus 和 AstraZeneca 协议包括向 Tempus 支付 $200M 数据许可和模型开发费用。这是真实经济承诺,但不能证明 Pathos 有流入收入。公开商业化动作因此应理解为业务拓展和治疗资产策略,而不是企业销售。牵引力的最佳外部证据,是公司能融资、签下重要交易对手,并通过 Rain Oncology 和 DeuterOncology 继续扩展管线。这是有意义的战略验证,但不等同于已验证的经常性商业收入。[CI014, CI015, CI016, CI017, CI018, CI019]
| 收入来源 | 机制 | 单位 | 当前状态 | 收入质量 | 尽调要求 |
|---|---|---|---|---|---|
| 资产对外授权 / 共同开发 | Pathos 控制的资产对外合作时,可收取首付款,以及后续开发和监管里程碑付款 | 按资产交易 | 未来可行的收入来源;Pathos 公开资料未披露交易经济条款 | 潜在毛利率较高,但目前未经验证 | 按资产索取已签署条款清单和历史 BD 管线 |
| 商业化资产特许权使用费 | 合作伙伴或收购方获批并上市后,按未来净销售额的一定比例分成 | 净销售额百分比 | 仅为未来选项;未披露特许权使用费表 | 若资产成功,长尾经济性有吸引力,但公开层面完全是假设 | 索取特许权使用费区间、阶梯下调和区域例外条款 |
| 临床开发合作 | 与制药对手方签共享开发或期权式协议 | 按合作项目 | 已披露合作关系,但未披露 Pathos 流入端经济条款 | 可能分散风险,但公开记录没有显示已实现现金流入 | 索取合作摘要,覆盖首付款、里程碑和成本分担条款 |
| AI / 试验设计服务 | 针对试验设计、生物标志物分析或组合分析收取服务费或软件费 | 按项目或订阅 | 没有公开定价、合同或收入证据 | 作为可变现业务线,目前还未获验证 | 索取 SOW、价卡和可计费客户数 |
| 内部资产价值兑现 | 通过出售、剥离或合作变现内部推进的资产 | 按交易 | 管线建设进行中;未披露公开退出 | 收入不平滑、由事件驱动,而非经常性 | 索取资产级变现计划和时间假设 |
| 基础模型经济性 | 未来可能围绕共享肿瘤学模型输出获得收入分成或许可 | 未说明 | 合作已存在;收入分成和所有权未披露 | 理论期权价值高,但今天没有公开经济条款 | 索取 IP 所有权矩阵和下游经济权利表 |
公开记录支持的是未来变现路径,而不是当前已确认收入。披露最清楚的经济条款是 Pathos 向 Tempus 支付的费用,而非 Pathos 的流入收入。
[CI014, CI016, CI017, CI019, CI020, CI021]| 变现要素 | 公开价格 / 单位 | 是否披露已实现经济性? | 实际公开内容 | 来源 / 限制 |
|---|---|---|---|---|
| 独立 PathOS 访问 | 未披露 | 否 | 未出现价目表、席位价格或合同模板 | 公开材料描述能力,但未给出商业条款 |
| AI 临床试验设计服务 | 未披露 | 否 | 未出现工作说明书定价或项目费披露 | Pathos 讨论试验设计价值主张,但没有收费条款 |
| 资产合作中应付 Pathos 的首付款经济条款 | 未披露 | 否 | 公开资料没有量化任何流入端首付款 | 收入侧合作经济条款仍未公开 |
| 应付 Pathos 的里程碑 / 特许权使用费 | 未披露 | 否 | 未出现里程碑阶梯或特许权使用费区间 | 公开记录证明存在可能性,不证明实际费率 |
| Tempus 基础模型协议 | $200M 对外付费承诺 | 是,但体现为成本而非收入 | 最具体的公开经济条款,是 Pathos 向 Tempus 付款 | 可用于判断成本结构;不能证明 Pathos 变现 |
| Schrödinger 混合模式(可比公司) | 软件收入加发现上行收益 | 是,仅作可比参照 | 上市公司可比案例展示了已披露混合模式的样子 | 可比公司经济性不是 Pathos 定价 |
| Insilico 同业收入披露(可比公司) | 2025 年收入 $56.2M | 是,仅作可比参照 | 同业展示了明确的 AI 生物技术收入披露应是什么样 | 可比公司经济性不代表 Pathos 收入 |
所有 Pathos 专属变现条目实际上都未披露。加入可比公司行,只是为了说明 Pathos 距离可公开建模的定价界面还有多远。
[CI015, CI016, CI020, CI032, CI043, CI047]从 Pathos 当前的资产与合作姿态,推导投资人需要看到哪些变现路径,才足以承保经常性收入质量。
Pathos 尚未按收入流披露已实现收入,因此本图是概念性梳理。它区分已披露的经济安排和未来可能的变现路径,并不暗示当前已经产生收入。
[CI014, CI015, CI016, CI017, CI020, CI021]4.3 成本结构、烧钱基准与单位经济
Pathos 公开单位经济数据基本不可得,因此本章采用生物技术优先的基准集,而不是强行套用 SaaS 指标。可能的成本桶很直接:pocenbrodib 持续临床执行,P-500 和 DO-2 等额外资产准备,与数据和模型开发绑定的付款,以及维持发现和临床开发组织的核心运营支出。肿瘤试验文献解释了为什么即便在商业化前,烧钱速度也可能维持高位。临床试验经济学综述强调,启动延迟、入组不足和基础设施碎片化会吞掉大量预算;公开肿瘤试验入组证据显示,入组放慢会直接拉长读出时间,从而延长融资窗口。 公开可比公司强化了资本强度这一点。Relay 披露 2025 年 Q1 现金约 $710M,现金跑道延至 2029 年,但依赖 ATM 增发后,2026 年 Q1 仍报告现金 $642.1M。Schrödinger 已有软件收入和合作发现经济收益,2026 年 Q1 仍报告现金 $406M、季度净亏损 $60M。Insilico 报告现金 $393.3M、2025 年收入 $56.2M,说明 AI 赋能发现公司即便建立了比 Pathos 更明确的商业收入流,也可能继续依赖资本。合起来看,这些基准支持一个合理公开烧钱区间:一家试图在肿瘤领域跑平台加管线模式的公司,年烧钱大约 $120M 至 $240M。在此基础上,Series D 单轮融资看起来可观,但并不过剩。[CI023, CI024, CI025, CI026, CI027, CI028]
| 指标 | 公开数值 / 估计 | 置信度 | 重要性 | 尽调要求 |
|---|---|---|---|---|
| 当前已确认收入 | 未披露 | 低 | 任何毛利率或倍数分析的起点 | 按收入来源索取自成立以来的月度收入桥 |
| 毛利率 | 未披露 | 低 | 决定该模式更像软件、服务重,还是生物技术 | 按收入来源索取毛利和分摊政策 |
| 年度烧钱基准 | 估计每年 ~$120M-$240M | 低 | 直接决定现金跑道和下一轮融资时间 | 提供 2024、2025 和 Q1 2026 实际净烧钱额 |
| Series D 后独立现金跑道 | 估计 ~18-36 个月 | 低 | 检验最新一轮融资能否撑到关键读出或下一次融资 | 提供交割现金、受限现金和已承诺但未支付义务 |
| 每项已披露资产对应资本 | 每项公开资产 ~$117M | 中 | 平台+管线公司的简易资本效率代理指标 | 按资产和平台工作流提供内部资本分配 |
| Series D 交割后在手现金 | 未披露 | 低 | 判断资本充足性最重要的缺失输入 | 提供已签署的交割声明和资金余额 |
| Tempus 费用承诺 | 已披露 $200M | 中 | 已知现金流出;具体付款时间可能显著压缩现金跑道 | 提供付款时间表和会计处理 |
| CAC / 回本周期 | 未披露 / 公开层面尚无意义 | 低 | 如果 Pathos 大规模销售软件或服务,这项指标会很重要 | 提供 BD 支出、合作伙伴漏斗转化和签约周期 |
| 销售效率代理指标 | 融资节奏和合作伙伴入口,而非 CAC | 中 | 商业漏斗缺失时,公开资料中最好的替代指标 | 提供对手方管线、周期长度和转化数据 |
| 营运资本 / 项目融资义务 | 公开资料未披露 | 中 | 隐性义务可能削弱看似强劲的现金头寸 | 提供债务明细、担保和不可取消承诺 |
公开资料下,对 Pathos 单位经济性的承销几乎被空白项主导。本表有意明确:哪些只能估计,哪些需要数据室证据。
[CI023, CI024, CI025, CI026, CI027, CI028]从已知公开经济安排和同业基准,搭出 Pathos 消耗现金与跑道的情景估算。
Pathos 未披露实际现金消耗、现金余额或毛利率。因此本图用公开同业和试验文献输入展示估算如何搭建,而不是声称精确。
[CI009, CI022, CI023, CI024, CI034, CI035]有来源支撑的区间,展示 Pathos 可能的现金消耗包络、情景跑道,以及其融资规模在更广义 AI 药物发现资本谱系中的位置。
低 / 中 / 高点使用公开可比公司和已披露融资规模。该区间用于定位,不替代管理层现金桥。
[CI026, CI029, CI031, CI033, CI034, CI039]4.4 估值背景与尽调阻断点
Pathos 披露的融资路径使其处于私有 AI 生物技术融资的上层,但尚未达到该类别顶端。按公开披露资本,Pathos 已融资约 $467M,显著高于 Variational AI $5.5M 种子轮延伸融资等较小 AI 发现融资,也高于 Insilico 报道的 $110M Series E,但仍远低于 Isomorphic Labs $2.1B Series B。更重要的比较是模型质量。Schrödinger 的官方战略把经常性软件收入与药物发现上行空间配对,而 Relay 披露则展示了更传统临床阶段治疗公司依靠大额现金储备和反复进入资本市场融资的经济逻辑。Pathos 在资本强度上更接近 Relay,而不是在当前收入多元化上接近 Schrödinger。 这一框架很重要,因为最关键阻断点不在于 Pathos 是否还能融一次钱,而在于它能否在再次融资前把资本转化为可衡量的商业或临床价值。Nature 的融资评论是反向提醒:即便复苏开始,生物技术融资环境也可能快速收紧。公司层面,Pathos 仍未披露手头现金、股权结构表条款、清算优先权、确认收入、利润率结构,或未来模型产出与伙伴之间的经济分成。Series D 投资者身份也大多未披露。结果是一家公司看起来资金充足且获得战略验证,但收入质量、真实现金跑道和利润率路径仍只能靠私有证据判断。[CI037, CI038, CI039, CI040, CI041, CI042]
| 缺失指标 | 重要性 | 当前最佳公开代理指标 | 精确尽调路径 | 优先级 |
|---|---|---|---|---|
| Series D 交割时在手现金 | 现金跑道和融资依赖的核心输入 | 仅有 Series D 规模和同业烧钱基准 | 索取资金报表、受限现金明细和交割资产负债表 | 关键 |
| 月度净烧钱及按职能拆分的烧钱 | 需要用它把资本与运营节奏连起来 | 同业基准区间为每年 ~$120M-$240M | 索取 24 个月现金桥,并拆分 R&D、G&A 和合作支出 | 关键 |
| 收入结构和收入确认政策 | 需要用它判断收入是经常性、里程碑式还是一次性 | 未披露公开收入项目 | 按收入来源索取经审计 P&L 和政策备忘录 | 关键 |
| 流入端合作经济条款 | 需要用它检验合作是产生现金流入,还是只有战略观感 | 公开披露的只有付给 Tempus 的流出费用 | 索取合同摘要,覆盖首付款、里程碑、特许权使用费和收入分成 | 关键 |
| 毛利率 / 单项资产成本 | 判断单位经济性和盈利路径必须要它 | 未公开毛利或资产成本 | 要求提供各项目毛利率桥接和资产级支出 | 关键 |
| Series D 投资方身份和证券条款 | 用于判断股东结构质量和后续融资信号 | 只用了“新老投资方”这类泛化表述 | 要求提供投资方名单、股份类别、每股价格和优先权结构 | 高 |
| 股权结构和持股集中度 | 用于评估稀释、治理和后续融资能力 | 未公开持股明细 | 要求提供完全稀释股权结构和董事会权利摘要 | 高 |
| 销售效率 / BD 漏斗指标 | 把合作方兴趣转成可融资的收入展望,需要这些指标 | 融资和合作方标识只能粗略代用 | 要求提供各阶段漏斗转化率、周期时长、管线和交易对手 | 中 |
| 资产级资本配置 | 用于判断平台是在提高资本效率,还是只是增加项目 | 4 项公开组合决策 / 资产,以及粗略的 $/asset 替代指标 | 要求按资产、平台、收购 / 授权类别拆分支出 | 中 |
这些不是边角料问题;要把叙事层面的信心推进到可建模的财务确信,最少也要拿到这些非上市公司披露。
[CI014, CI015, CI018, CI021, CI035, CI036]4.5 展项
05产品与技术
5.1 平台架构:PathOS 是运营模型,不只是单一模型
Pathos 公开材料持续把公司描述成不只是拥有一个领先资产的传统生物技术公司。核心产品主张是名为 PathOS 的运营模型,它围绕共同肿瘤数据层和湿实验室反馈闭环,整合 Foundry、Scout、Sprint 三个命名引擎。Foundry 被定位为共享 AI 核心,跨项目累积学习;Scout 从多模态证据中优先排序资产和患者亚群;Sprint 则是执行层,把选定资产从一个临床里程碑推向下一个。这对尽调很重要,因为技术故事和资产故事不可分割:Pathos 要求投资者相信,同一套数据和决策系统可以反复寻找、筛选并推进外部肿瘤资产,效果优于传统生物技术工作流。 公开记录中最强的证明,是模块描述足够具体,而且官网首页、平台页和管线页的运营模型叙事一致。Pathos 声称可访问超过 200 PB 多模态肿瘤数据,并将其与生物标志物发现、患者选择和试验设计配对。外部 Tempus 合作提供了重要的第二见证:Tempus 称其去标识化肿瘤数据将用于构建共享基础模型,而 Pathos 称完成后的模型将在三方之间共享。因此,下方产品架构图和模块矩阵把 PathOS 视为一套整合栈,依赖外部数据、内部 AI 决策和资产级执行,而不是独立软件 SKU。[CE001, CE002, CE003, CE004, CE005, CE006]
| 模块 / 资产 | 主要用户 | 当前状态 / 成熟度 | 作用 | 差异化主张 | 尽调缺口 |
|---|---|---|---|---|---|
| Foundry | Pathos 管理层和平台团队 | 早期成形 / 内部证据支撑 | 共享 AI 核心,负责资产排序、组合决策,并把实验室反馈接回学习闭环 | 主张能跨项目学习,并用 AI 原生方式筛选资产组合 | 未公开准确率或决策质量指标 |
| Scout | 发现 / 转化团队 | 已公开描述,但没有外部基准 | 借助多模态证据为资产和响应患者亚群排序 | 在资产收购前把机制与亚群选择连起来 | 未公开与传统 BD 流程对比的基准 |
| Sprint | 临床开发团队 | 已用于活跃项目 | 推动筛选后的资产进入生物标志物驱动的临床执行 | 由小型 AI 赋能小队压缩决策周期 | 未公开周期时长或每个里程碑成本数据 |
| Pocenbrodib (P-300) | 临床运营和肿瘤学研究者 | Pathos 赞助的 phase 1b/2a 正在进行 | mCRPC 方向的核心 CBP/p300 抑制剂项目 | 把既有 FT-7051 数据与 Pathos 生物标志物策略结合 | Pathos 主导研究尚未公开疗效读出 |
| P-500 / PRT811 | 转化肿瘤团队 | 既有 phase 1 之后,处于中期前规划 | 面向高级别胶质瘤 / 葡萄膜黑色素瘤的可入脑 PRMT5 抑制剂 | 既有 phase 1 信号,加上 Pathos 的亚群选择论点 | 尚无 Pathos 生成的临床数据 |
| DO-2 | 组合 / 临床策略 | 2026 年新近收购 | 通过 Foundry 找到的第三代 MET 抑制剂 | 早期 METex14 响应强,且主张在水肿上有差异化 | 整合和复现早期信号仍待验证 |
| Milademetan / Rain 资产 | 公司发展 | 已持有,但公开叙事中降调 | Rain 收购带来的存量精准肿瘤项目 | 把自有资产基础扩到两个头部项目之外 | 近期 Pathos 材料看不清当前优先级 |
结合 Pathos 官方披露和独立报道;成熟度指公开证据强度,不代表内部准备度。
[CE001, CE004, CE005, CE017, CE020, CE024]| 层级 / 组件 | 角色 | 证据 | 关键依赖 | 主要风险 |
|---|---|---|---|---|
| 外部多模态肿瘤数据 | 模型训练和亚群发现的底层数据 | Pathos 声称 >200 PB;Tempus 拥有 8.5M+ 条记录和 450+ PB | Tempus 数据权利和持续合作 | 数据权利、隐私和集中度风险 |
| Foundry 核心 | 汇总跨项目学习,并提出组合决策 | Pathos 平台页和 Deuter 收购公告 | 内部模型治理和湿实验室联动 | 未公开验证基准 |
| Scout 引擎 | 为资产和患者人群排序 | Pathos 平台页和 Series C 说明 | 因果模型和标注结局的质量 | 虚假亚群选择风险 |
| Sprint 执行小队 | 把选定资产转化为试验运营 | Pathos 平台页 | 临床运营执行和合作方支持 | 难以把软件杠杆与团队质量拆开 |
| 湿实验室验证闭环 | 测试并细化模型得出的洞察 | Pathos 平台页 | 实验室吞吐量和实验设计 | 公开覆盖范围和节奏未披露 |
| 外部试验赋能网络 | 加快活跃研究的启动和入组 | Tempus TIME 案例研究 | Tempus 网络可用性 | 合作方激励变化会带来执行风险 |
架构仅反映公开描述;若干关键控制点仍只有私下证据。
[CE002, CE005, CE013, CE015, CE016, CE035]PathOS 围绕具体资产,叠加外部肿瘤数据、共享 AI 核心、决策引擎和试验执行。
[CE001, CE002, CE005, CE012, CE013, CE024]5.2 公开资产图谱:Pathos 正用许可和收购的肿瘤项目验证系统
公开管线狭窄但具体。Pathos 在管线页列出 pocenbrodib 和 P-500,2026 年公告又通过收购 DeuterOncology 加入 DO-2,并在 2024 年要约收购完成后保留 Rain 的 milademetan 遗留所有权。Pocenbrodib 是最清晰的运营证明,因为 Pathos 已将其推进到由 Pathos 赞助的转移性去势抵抗性前列腺癌 1b/2a 期研究。公司称该研究结合单药和三个联合用药臂,目标约 203 名患者,并直接建立在早期 FT-7051/COURAGE 经验之上。Tempus TIME 案例研究提供了另一项运营证明,显示 Pathos 如何借助外部试验赋能网络识别前 10 名入组患者中的 6 名。 P-500 和 DO-2 很重要,因为它们展示了公司的收购论点。Pathos 在此前 1 期工作显示 IDH 阳性高级别胶质瘤信号后许可引进 P-500;DO-2 则由 Pathos 于 2026 年收购,此前 Foundry 据称基于机制、药代动力学和竞争定位将其提升为优先候选。这些资产不是随机新增;Pathos 明确声称,平台能够发现低估的临床阶段药物,将它们重新映射到更合适的生物标志物定义人群,再沿更数据驱动的开发路径推进。这个主张值得跟踪,但距离端到端证明仍很远,因为公开证据偏重交易和试验启动里程碑,闭环绩效指标很少。[CE017, CE018, CE019, CE020, CE021, CE022]
| 工作流步骤 | 当前工作流 | Pathos 方案 | 公开引用的可衡量收益 | 主要限制 |
|---|---|---|---|---|
| 资产寻源 | 传统生物技术 BD 依赖人脉和会议 | Foundry 筛查临床和科研数据,寻找被低估资产 | Deuter 交易被包装为 Foundry 寻源的组合决策 | 没有独立审计其寻源精度 |
| 患者亚群选择 | 试验人群常靠粗颗粒入排标准筛选 | Scout 将机制与多模态响应和耐药信号连起来 | Pathos 称 P-500 和 pocenbrodib 策略采用生物标志物驱动的亚群划分 | 未公开前瞻性命中率统计 |
| 试验设计 | 标准阶段推进依赖较慢的人工综合判断 | Sprint 小队嵌入 AI 科学家和工程师,把资产从一个里程碑推向下一个 | Pathos 称每项试验都会借助数据反馈变得更“聪明” | 未披露相较同业试验的周期差异 |
| 入组加速 | 早期肿瘤研究常卡在中心启动慢、患者匹配慢 | 在 pocenbrodib 研究首批 10 个匹配中,Tempus TIME 网络识别出 6 个 | 有具名外部证据表明 Pathos 能接入规模化入组基础设施 | 收益目前依赖外部网络 |
| 实验室验证 | 很多 AI 药物发现叙事止步于模型输出 | Foundry 主张用实验室在环验证来检验并复用洞察 | 从概念上强化学习飞轮 | 未公开检测吞吐量和验证规则 |
收益只限 Pathos 或 Tempus 公开材料已披露的主张;KPI 未披露本身被视作尽调缺口。
[CE005, CE010, CE024, CE025, CE033, CE035]Pathos 把药物开发描述为一条可重复序列:从数据和外部资产获取,走到生物标志物指导下的试验执行。
[CE003, CE004, CE006, CE010, CE024, CE033]5.3 依赖与控制:最大的技术杠杆伴随外部数据、隐私和治理风险
Pathos 的技术优势,也是它最大的依赖栈。Tempus 和 AstraZeneca 合作具有战略吸引力,因为它把 Pathos 扩展到自身内部数据足迹之外,但也带来实质交易对手风险。公开看,Pathos 依赖 Tempus 提供去标识化肿瘤数据、企业级多模态基础设施,并且——借 TIME Network 案例研究可见——甚至获得可衡量的试验入组帮助。这种集中不必然是坏事,但意味着从模型到临床执行的路径并非完全由 Pathos 独自控制。如果合作经济条款、数据权利或交付优先级变化,公开技术论点也会随之变化。 已审阅公开材料也没有达到机构投资者理想中应有的治理深度。HHS 提醒受监管实体,即便妥善去标识化的健康数据仍有一定残余再识别风险;NCI 和多篇肿瘤 AI 综述强调,生物标志物工作流、同意、隐私和模型可靠性仍是规模化采用的不小障碍。The ASCO Post 和 Springer 综述尤其切中要害,因为它们追问责任、同意、可解释性和训练数据质量——当平台影响资产筛选或试验设计时,这些类别会变得尖锐。Pathos 的招聘页面显示技术招聘和基础反欺诈控制,但公司未公开信任中心、运行时间历史、公开模型卡或已验证平台 KPI 仪表盘。对尽调而言,控制体系叙事方向上可行,但公开文档仍不足。[CE011, CE012, CE013, CE014, CE015, CE016]
| 控制项或问题 | 当前公开状态 | 范围 | 证据显示什么 | 缺口 |
|---|---|---|---|---|
| HIPAA 去标识化 | 框架存在,但仍有残余风险 | 任何用去标识化、患者关联数据训练的模型 | HHS 称 Safe Harbor 和 Expert Determination 都仍有一定再识别风险 | 需要 Pathos 专属去标识化方法和监控细节 |
| 生物标志物有效性 | 临床上重要,但并非普遍常规 | 患者选择和伴随诊断式工作流 | NCI 称生物标志物检测是精准医疗核心,但对多数患者还不是常规 | 需要证据证明 Pathos 亚群逻辑能转化为真实入组和结局 |
| AI 责任和知情同意 | 监管 / 法律标准仍在演进 | 任何 AI 辅助肿瘤决策支持 | ASCO 和 Springer 来源强调责任、可解释性、偏差和知情同意问题 | 需要模型卡、监督结构和临床医生责任映射 |
| 安全 / 信任中心 | 已审阅 Pathos 材料中未出现公开信任门户 | 数据合作方和试验中心的企业级尽调 | Pathos 招聘页展示的是招聘控制和反欺诈提示,不是平台保障 | 需要认证、事件历史和变更控制文档 |
| 平台输出质量体系 | 未公开描述 | 输出用于试验设计或组合决策 | 公开叙事对愿景讲得细,对 QA 阈值讲得少 | 需要内部验证 SOP 和发布治理 |
这是控制就绪度视角,不是声称 Pathos 不合规;文档缺失本身就是主要尽调问题。
[CE041, CE042, CE043, CE044]公开可见的 Pathos 技术栈高度依赖 Tempus 的数据和试验赋能,也依赖公共卫生规则来实现肿瘤数据的隐私安全复用。
[CE011, CE013, CE014, CE041, CE045, CE046]5.4 成熟度评估:Pathos 已从概念走到试验运营,但平台证据仍不完整
成熟度图景并不单一,恰是成熟投资人更该喜欢的状态:公开证据足以说明 Pathos 在真做事,但还不足以证明平台优势已经跑通。正面看,Pathos 已经授权并重新命名首个临床资产,启动公司主导的 1b/2a 期试验,拿到第二个带有既往 1 期信号的临床资产,完成 Rain 收购,并通过控股收购拿下 DO-2,同时连续完成 Series C 和 Series D 融资。这串动作比一般「AI 药物发现」故事更具运营实感。 悬而未决的是:PathOS 到底是可复制的生产力引擎,还是只是套在标准生物科技资产套利和执行流程外的一层有用工具。Pathos 尚未公开资产排序精度、生物标志物选择命中率、试验周期压缩、验证吞吐量或模块级可靠性等指标。因此,模块成熟度矩阵把资产交易和试验启动层标为已有证据,把基础模型建设标为正在成形,把平台治理层标为披露不足。实际承销立场是:Pathos 的产品成熟度已经足以进入严肃尽调,但若没有私有数据室证据,公开指标化证据还不足以把系统按已验证的复利引擎承销。[CE020, CE024, CE025, CE033, CE034, CE035]
| 日期 / 时期 | 里程碑 | 状态 | 含义 | 来源 |
|---|---|---|---|---|
| 2023-10 | 从 Novo Nordisk 授权引进 FT-7051 / P-300 | 已完成 | 让 Pathos 进入临床阶段资产持有 | Pathos 官方 |
| 2024-01 | 完成 Rain Oncology 收购 | 已完成 | 增加包括 milademetan 在内的全资肿瘤项目 | Pathos 官方 |
| 2024-08 | 从 Prelude 生态授权引进 PRT811 / P-500 | 已完成 | 增加第二个有既往 phase 1 数据的差异化临床阶段资产 | BioSpace |
| 2024-10 | 完成 $62M Series C 融资 | 已完成 | 为团队扩张、平台扩展和 P-300 / P-500 推进提供资金 | Pathos 官方 |
| 2025-03 | 首例 pocenbrodib 患者给药 | 里程碑已完成 | 建立 Pathos 赞助的临床执行记录 | Pathos / Urology Times / Tempus |
| 2025-04 | 签署 Tempus 与 AstraZeneca 基础模型合作 | 已完成 | 放大数据和模型野心,但也提高合作方依赖 | Pathos / Tempus |
| 2025-05 | 完成 $365M Series D,投后估值约 $1.6B | 已完成 | 为管线和 AI 建设延长资金储备 | Pathos / 独立报道 |
| 2026-05 | 收购 DeuterOncology / DO-2 多数股权 | 已完成 | 显示 Foundry 主导的外部资产寻源仍在推进 | Pathos / BioSpace |
公开路线图交易色彩很重;下一步真正证据是 Pathos 主导研究披露疗效和运营 KPI。
[CE017, CE020, CE024, CE028, CE034, CE039]公开证据在交易和试验启动上最强,在可衡量的平台表现和治理上较弱。
[CE020, CE024, CE033, CE040, CE041]5.5 图表
06客户情况
6.1 客户可见度:公开证据显示的是企业级关系,不是广泛披露的客户基础
从客户披露角度看,Pathos 不像传统软件公司或诊断公司。官网和融资材料没有披露客户数、ARR、留存或定价。公司转而把自己描述为与制药公司、生物科技公司、学术界和投资者合作,搭建新的肿瘤药物开发运营模型。这个定位很关键:它意味着短期买方更可能是企业级交易对手——战略制药合作者、数据网络或试验赋能伙伴——而不是大批购买标准化产品的肿瘤医生或医院。因此,客户细分表把 Pathos 已命名的外部关系拆成联合开发伙伴、数据 / 试验赋能供应商、资产交易对手和临床用户。 直接结论是,公开客户可见度仍低,尽管公开关系证据是真实的。Pathos 有足够证据证明外部方愿意与其合作,但不足以勾勒多元化客户账本。这与 Tempus、Flatiron、Foundation Medicine、Caris 和 ConcertAI 形成鲜明对比,后者都披露了更强的客户或使用规模说法。尽调中应把 Pathos 视为由关系驱动的生物科技平台,外部需求信号正在出现,而不是已经透明规模化的商业平台。[CU001, CU002, CU003, CU013, CU017, CU018]
| 细分 | 买方 / 用户 / 付款方 | 当前公开证据 | 规模信号 | 收入 / 战略价值 | 缺口 |
|---|---|---|---|---|---|
| 战略共同开发合作方 | 买方:药企合作方;用户:R&D / 数据科学;付款方:合作预算 | AstraZeneca + Tempus 基础模型合作 | 有具名披露,但经济条款不透明 | 若可重复,战略价值高 | 未披露归属 Pathos 的收入分成 |
| 数据 / 试验赋能供应商 | 买方:Pathos;用户:临床运营和转化团队;付款方:Pathos 预算 | Tempus 数据 + TIME 网络案例研究 | 首批 10 名匹配患者中 6 名来自 TIME | 运营价值高;可能是经常性支出 | 更显示 Pathos 是客户,而不是供应商 |
| 临床研究者 / 中心 | 用户:试验中心;付款方:研究赞助方 | pocenbrodib 研究有 3 个美国中心 | 真实部署,但覆盖面小 | 对试验执行重要 | 未公开中心名单或扩张趋势 |
| 资产交易对手 | 买方 / 卖方随交易而变 | Novo、Prelude、Rain、Deuter 交易 | 多笔具名交易 | 战略管线价值 | 不是经常性客户证明 |
| 未来平台买方 | 可能是寻求患者数据和试验设计杠杆的药企 / biotech 项目 | 根据定位和现有关系推断 | 未公开量化 | 若验证成立,潜在空间大 | 未披露客户数、定价或漏斗 |
细分表区分买方、用户和交易对手;Pathos 公开披露会把这些类别揉在一起,因此这里明确拆开。
[CU002, CU003, CU004, CU007, CU012, CU025]Pathos 当前公开客户旅程从战略引荐和资产交易走到现场试验执行,但仍缺少透明的续约指标。
[CU003, CU007, CU008, CU011, CU031, CU038]6.2 具名外部证据:Tempus 是真实采用最清晰的证据,但主要证明 Pathos 是买方
最强的外部具名证据是与 Tempus 的关系。第一,Tempus 与 AstraZeneca 的合作显示,成熟交易对手愿意与 Pathos 一起搭建肿瘤基础模型。第二,Tempus TIME 案例提供运营证据:Pathos 正在借第三方网络推进早期试验。全章最具体的数据点是,pocenbrodib 研究前 10 名匹配患者中有 6 名来自 TIME 网络。Tempus 还引用 Iker Huerga 的说法,说明该网络帮助 Pathos 找到可能有合格患者的研究中心并快速启动。合在一起,这些事实把 Pathos 从概念性 AI 故事推进到有记录的企业级使用。 但同一组证据也暴露重要不对称:公开证据最清楚显示的,是 Pathos 作为 Tempus 数据、网络和商业化基础设施的客户。因此,具名客户证据表把 Tempus 同时视为 Pathos 最强外部验证者和最大商业依赖。AstraZeneca 提供战略验证,但公开记录仍很少说明 Pathos 到底直接掌握多少经济权益或工作流。因此,下方采用漏斗把当前采用进展解读为申办方级企业证据,而不是广泛平台渗透。[CU004, CU005, CU006, CU007, CU008, CU009]
| 指标 / 里程碑 | 数值 | 日期 | 来源 | 置信度 | 含义 | 缺失分母 |
|---|---|---|---|---|---|---|
| 具名战略合作 | 宣布 Tempus + AstraZeneca | 2025-04-23 | Pathos + Tempus 新闻稿 | 高 | 证明大型企业兴趣真实 | 未披露 Pathos 收入贡献 |
| 数据授权 / 模型开发经济性 | Tempus 将获得 $200M 费用 | 2025-04-23 | Pathos + Tempus 新闻稿 | 高 | 显示 Pathos 愿意为外部平台投入付费 | 未披露 Pathos 侧预算 / ROI |
| Pocenbrodib 研究入组证明 | 迄今匹配 10 名患者,其中 6 名来自 TIME | 2026-02 PDF / 2025-03 试验背景 | Tempus TIME 案例研究 | 高 | 最强公开实时采用数据点 | 未披露完整入组漏斗或筛选失败率 |
| 临床部署覆盖面 | 3 个美国试验中心 | 2025-03-20 | Urology Times | 高 | 真实,但运营覆盖面仍窄 | 未见中心扩张趋势 |
| 客户数 / ARR / NRR | null | 2026-05-25 | 公开材料中未见披露 | 中 | 商业透明度弱 | 分母完全缺失 |
空值表示截至运行日期没有公开披露,不代表活动为零。
[CU004, CU005, CU008, CU011, CU028]| 交易对手 | 细分 | 部署 / 用例 | 生产使用还是试点 | 结果 / 证据 | 限制 |
|---|---|---|---|---|---|
| Tempus AI | 供应商 + 客户验证 | 数据授权、基础模型构建、TIME 试验网络 | 生产环境 / 在研试验支持 | pocenbrodib 研究前 10 名匹配患者中有 6 名来自 TIME | 说明 Pathos 是 Tempus 基础设施的买方 |
| AstraZeneca | 战略药企合作方 | 围绕肿瘤学基础模型的多年共同开发 | 生产级合作 | 具名共建方,也是未来模型用户 | 未披露 Pathos 经济条款或扩张计划 |
| Novo Nordisk / Forma 来源 | 资产交易对手方 | 将 FT-7051 / P-300 授权纳入 Pathos 管线 | 已完成交易 | 让 Pathos 拿到首个临床阶段资产 | 一次性交易,不能证明持续客户需求 |
| Prelude / P-500 来源 | 资产交易对手方 | 授权带有既往 1 期数据的 PRMT5 项目 | 已完成交易 | 扩充了 Pathos 的临床阶段资产组合 | 仍只证明能做交易,不证明重复使用 |
公共材料中可见的具名外部证据只是不完全枚举;表格刻意把真实客户验证和交易对手方放在一起,因为公开证据稀疏到只能这样呈现。
[CU004, CU007, CU008, CU025, CU026, CU027]公开采用漏斗很窄:从关系新闻走到少数具名现场证明点。
[CU004, CU007, CU008, CU011, CU028, CU039]6.3 持久性与留存:公开记录证明关系存在,但不能证明会留下或扩张
这是公开客户故事中最弱的一环。Pathos 没有披露 NRR、GRR、流失、合同期限、客户满意度或扩张收入。也没有定价或包装细节,无法让投资者推断账户如何从试点走向持续支出。这意味着本章能验证采用事件,却不能验证持久性。留存表刻意保留大量空值,因为这才是公开证据的真实状态。 缺失数据对 Pathos 比对成熟同行更重要,因为少数已命名关系容易制造虚假信心。一个合作标题可能代表长期项目、短期工作说明书,或一次性交易。同样问题也适用于 Novo Nordisk、Prelude、Rain 和 Deuter 等资产交易对手:这些交易证明 Pathos 能做交易,但不能证明平台有重复商业需求。在 Pathos 披露续约、重复范围扩张或客户提供的高质量参考成果之前,正确分析立场是:公开证据尚未证明关系持久性。[CU027, CU028, CU029, CU030, CU035, CU038]
| 指标 | 数值 | 分组 | 置信度 | 尽调索取 |
|---|---|---|---|---|
| 续约率 | null | Pathos 所有关系 | 低 | 索取续约时间表和工作说明书历史 |
| NRR / GRR | null | 任何经常性企业账户 | 低 | 按账户索取分队列收入桥 |
| 合同期限 | null | Tempus / AstraZeneca / 其他 | 低 | 索取初始期限和续约机制 |
| 客户满意度 / 推荐证明 | null | 具名交易对手方 | 低 | 索取直接客户访谈和实施复盘 |
| 扩张证据 | 一项研究中前 10 个 TIME 匹配有 6 个 | 仅 Tempus 关系 | 中 | 需要证明它能跨项目扩张,而不只是一项试验 |
截至 2026-05-25,null 表示没有公开披露;唯一正面数据点是试验运营支持,不是留存。
[CU008, CU028, CU029, CU035]公开证明质量在关系存在上最强,在持久变现上最弱。
[CU004, CU007, CU008, CU025, CU029, CU039]6.4 集中度与扩张:Pathos 可能赢得更多伙伴,但当下公开客户图景很窄
集中度风险很直接。Tempus 在公开证据中反复出现:数据供应商、模型建设合作者、试验赋能网络,以及最清晰的客户证据式材料来源。AstraZeneca 是另一个重要的具名战略交易对手。除此之外,公开故事主要是一组交易或自有资产,而不是多元化付费用户名单。Pathos 是私营公司,又不披露收入分成细节,公开证据无法按美元或账户量化集中度;只能定性说明集中。 扩张仍然有可能。Pathos 现在已有一项正在进行的申办方试验、第二个带有既往 1 期信号的资产、一个新的 MET 项目和刚拿到的 Series D 资金。如果公司能证明其生物标志物逻辑提升入组或响应概率,下一步自然是更多制药公司合作和更多申办方级研究。但市场挤满了数据披露更充分的肿瘤平台。Tempus、Caris、Foundation Medicine、Flatiron、ConcertAI、PathAI、Owkin 和 Recursion 都披露了更清晰的商业或使用信号。Pathos 仍可能赢,但必须先把具名证据转化成更广、更可测的客户基础,本章才能成为明确优势。[CU012, CU013, CU014, CU015, CU016, CU020]
| 扩张驱动因素 | 集中度风险 | 影响 | 尽调路径 |
|---|---|---|---|
| 更多申办方级研究 | 现有证据锚定 Tempus 和 AstraZeneca | 任一关系变弱,Pathos 都会失去可见客户证据 | 索取合作方管线和备选供应商 |
| 更广泛的药企合作 | 除少数名称外,没有公开客户名单 | 可能压住估值倍数扩张和增长信心 | 索取完整 BD 漏斗和活跃机会 |
| 生物标志物驱动商业化 | 生物标志物检测并非处处常规化 | 可能拖慢向医疗服务方或支付方放量 | 按照照护场景索取采用率假设 |
| 数据网络撬动 | 去标识化数据仍背着治理负担 | 可能拖慢采购或合同扩张 | 审阅隐私控制和数据权利文件包 |
| 竞争差异化 | 工具链成熟的同业已经在服务肿瘤买方 | 缺少更清晰证据,Pathos 可能难以赢得大范围预算 | 用具名竞争对手压测赢单 / 输单数据 |
集中度只能定性判断,因为公开证据不足以测算收入占比。
[CU024, CU025, CU031, CU032, CU033, CU034]由于公开队列数据缺失,读者应把耐久性视为一条未解决的尽调路径,而不是已被证明的增长引擎。
[CU028, CU029, CU035, CU040]6.5 图表
07风险
7.1 风险概览:Pathos 面临集中的伙伴、治理和早期临床风险
Pathos 的风险还不是晚期生物科技公司的那种风险;它的风险来自一个高度集中的 AI 原生肿瘤运营模型。公司试图把外部数据权利、模型建设、生物标志物逻辑、资产交易和试验执行拼成一个自我改进系统。这个结构带来上行空间,也把多个失效模式集中在同一条运营链上。如果伙伴数据权利收窄、生物标志物逻辑无法泛化、试验入组停滞,或治理标准不达标,系统可能在收入或关键疗效到来之前就在多个点断裂。因此,风险热力图把数据权利集中、法律和同意不确定性、早期临床执行和融资依赖放在登记册高位。 这不是纯理论担忧。公开记录已经显示:Pathos 严重依赖 Tempus 的数据规模和试验赋能;公司有多个收购或授权资产,仍需要整合和排序;关于 PathOS 输出在影响决策前如何验证,公开治理细节很少。公司仍是私营,投资者拿不到上市精准肿瘤公司通常会提供的风险因子密度。正确姿态是:除非私有尽调材料证明相反,否则假设剩余风险仍高。[CR001, CR002, CR004, CR005, CR012, CR013]
Pathos 最重要的风险,大多落在中高发生概率和高到严重影响区间。
[CR001, CR006, CR007, CR014, CR024, CR032]7.2 法律、监管和数据风险始于隐私、同意和未定型的 AI 责任
法律和监管栈是结构上最重要的风险区域,因为它贯穿 Pathos 的每个项目。HHS 明确指出,去标识化健康数据并非没有风险;即使正确遵循隐私规则,残余再识别风险仍然存在。这个问题很关键,因为 Pathos 的公开逻辑依赖大规模患者关联肿瘤数据池,以及用 Tempus 数据构建的基础模型。如果监管者、交易对手或企业客户对现有去标识化做法的舒适度下降,Pathos 核心训练和证据层可能变得更贵、更慢或受限更多。 同时,肿瘤专属 AI 责任仍未定型。The ASCO Post 把 AI 指导诊断和治疗的责任标准描述为一个演进中的问题;Springer 综述则提示自动化偏见、知情同意、可解释性,以及训练数据本身的偏差。NCI 又增加了一项采用约束:生物标志物检测很重要,但对大多数患者而言并非常规流程,这意味着 Pathos 以生物标志物为核心的逻辑仍会碰到真实世界工作流摩擦。FDA 的 DTC 指南并不直接针对 Pathos,但强化了一个总体结论:数据驱动健康工具的证据质量和监督差异很大。结果是:法律和监管风险不是 Pathos 的边缘问题,而是核心商业模式的一部分。[CR002, CR003, CR004, CR006, CR007, CR008]
| 风险 | 司法辖区 / 规则体系 | 当前状态 | 可能性 | 严重度 | 缓释措施 | 剩余敞口 | 尽调路径 |
|---|---|---|---|---|---|---|---|
| 去标识化数据的残余再识别风险 | HIPAA / HHS | 结构性类别风险 | 中 | 高 | 合同控制、专家判定、访问控制 | 高 | 审阅 Pathos-Tempus 数据权利和去标识化文件包 |
| 决策支持中的 AI 责任 | 美国侵权 / 产品责任框架 | 尚未定型,仍在演变 | 中 | 高 | 人工复核,并记录 AI 影响 | 高 | 索取模型治理和责任分工图 |
| 偏见 / 同意 / 自动化偏见 | 临床伦理 / 肿瘤工作流 | 近期文献正在密集讨论 | 中 | 高 | 知情同意措辞、人工监督、偏见监测 | 高 | 索取试验知情同意和偏见监测材料 |
| 生物标志物采用摩擦 | 临床实践 / 支付方工作流 | 重要,但并非所有场景都已常规化 | 中 | 中 | 使用经过清楚验证的生物标志物逻辑,并培训站点 | 中 | 索取真实世界采用假设和站点准备度 |
| 健康工具证据质量错配 | FDA 器械 / 检测监管类比 | 数据驱动工具的证据质量参差不齐 | 中 | 中 | 验证,以及清晰的声明边界 | 中 | 对照验证文件包审阅 Pathos 声明 |
严重度按投资人尽调视角排序,不构成法律意见。
[CR002, CR003, CR006, CR007, CR008, CR009]7.3 运营和伙伴风险被放大,因为最干净的公开证据都经过 Tempus
公开证据显示 Pathos 能执行,也显示这种执行多依赖外部基础设施。Tempus TIME 案例之所以有价值,恰恰因为它很具体:早期研究会遇到研究中心和患者瓶颈,而 Tempus 称其帮助 Pathos 研究识别了前 10 名匹配患者中的 6 名。这令人鼓舞,但也意味着 Pathos 一个核心运营证明点位于 Pathos 之外。同样,Tempus 与 AstraZeneca 的合作给了 Pathos 难以独自搭建的规模,也让伙伴连续性变成现实风险变量,而不是背景细节。 临床风险同样不轻。Pocenbrodib 还必须跨过早期安全性和疗效阈值,才能进入更成熟的价值叙事。P-500 有早期信号潜力,也有明确不良事件历史;DO-2 则是刚收购的资产,基于一个小型 1 期数据集,仍需在更广人群中独立确认。叠加 Rain 整合和组合优先级问题,运营图景变成:确有进展,但韧性有限。下方依赖关系图把 Tempus 视为当前 Pathos 风险栈中最重要的外部节点。[CR012, CR013, CR014, CR015, CR016, CR017]
| 失效模式 | 可能性 | 严重度 | 缓释成熟度 | 剩余敞口 | 未解决缺口 |
|---|---|---|---|---|---|
| 入组瓶颈或研究启动延误 | 中 | 高 | 有合作方支持,但尚未完全内化 | 高 | 可见公开证据依赖 TIME 网络 |
| Pocenbrodib 未达到进一步扩张所需的安全性 / 疗效门槛 | 中 | 严重 | 只有早期临床控制 | 严重 | 尚无 Pathos 自己生成的疗效读出 |
| P-500 不良事件谱限制后续开发 | 中 | 高 | 既往 1 期已有信息,但仍处早期 | 高 | 需要 Pathos 专属试验设计和安全策略 |
| DO-2 早期信号无法在更广人群中复现 | 中 | 高 | 收购后阶段很早 | 高 | 需要小型 1 期数据集之外的外部验证 |
| 平台质量 / 事件控制披露不足 | 中 | 高 | 公开证据无法判断 | 高 | 没有公开正常运行时间、CAPA 或事件历史 |
运营缓释看起来真实存在,但公开记录仍然太薄,无法把任何一项评为低剩余敞口。
[CR012, CR013, CR014, CR015, CR017, CR018]| 依赖项 | 交易对手方 | 角色 | 集中度 | 失效情景 | 严重度 | 缓释措施 | 剩余敞口 |
|---|---|---|---|---|---|---|---|
| 基础模型数据供给 | Tempus | 数据提供方和模型构建合作方 | 高 | 数据访问收窄或商业条款恶化 | 严重 | 备选供应商和合同保护 | 高 |
| 试验匹配 / 启动 | Tempus TIME 网络 | 入组支持 | 高 | 站点启动或匹配支持变弱 | 高 | 随时间推移内化更多站点运营 | 高 |
| 共同开发验证 | AstraZeneca | 战略外部验证方 | 中 | 合作范围不扩张,或势头减弱 | 中 | 建立更多元的药企关系 | 中 |
| 收购资产整合 | Rain / Deuter 资产 | 通过 M&A 扩大资产组合 | 中 | 整合分散管理层注意力,或掩盖挫折 | 高 | 明确资产组合审查节奏和阶段关口 | 高 |
| 供应商规模不对称 | 大型外部数据 / 诊断同业 | 企业级预期参照点 | 中 | Pathos 治理落后于合作方要求 | 高 | 加强控制披露和 QA 系统 | 高 |
Tempus 出现在多行,是因为公开证据反复经由这个合作方。
[CR004, CR005, CR013, CR022, CR023, CR030]| 角色 / 职能 | 依赖或缺口 | 可能性 | 严重度 | 缓释措施 | 尽调路径 |
|---|---|---|---|---|---|
| 平台治理 | 公开控制文件稀疏 | 中 | 高 | 正式审查委员会和模型卡 | 索取治理文件包 |
| 资产组合优先级 | 部分收购资产被弱化但原因不透明 | 中 | 高 | 阶段关口式资产组合审查 | 索取当前资产组合演示材料 |
| 临床领导层带宽 | 多个资产与模型构建同时推进 | 中 | 中 | 专门项目负责人 | 按项目索取组织架构图 |
| 隐私 / 法务领导层可见度 | 同业公开的专门控制负责人比 Pathos 更多 | 中 | 中 | 面向外部的信任与治理文件 | 索取具名负责人和审查节奏 |
| 资本配置纪律 | 多轮融资后仍缺少公开收入透明度 | 中 | 高 | 基于里程碑的支出控制 | 按项目索取预算和止损规则 |
本表关注执行结构和可见度,不评价个别高管。
[CR022, CR023, CR024, CR032, CR033, CR038]公开风险集中在少数外部合作方、监管机构和内部治理未知项上。
[CR004, CR005, CR012, CR013, CR024, CR036]7.4 财务和市场风险只有在 Pathos 持续把资本转化为可信证明点时才可控
Pathos 显然能融资,但这不等于融资风险低。SEC 提交材料显示,公司在 2023、2024 和 2025 年反复提交 Form D,最终形成 2025 年约 $400 million 的发行。一方面,融资历史是优势——投资者反复支持公司;另一方面,它也证明该模型仍依赖外部资本,而不是公开收入。只要公司仍处于商业化前、客户变现不透明,每一次未来融资都高度依赖市场继续相信平台叙事,以及新的临床或运营证明。 品类风险会叠加公司自身故事。Tempus 和 Caris 表明,规模化肿瘤数据公司同样要处理隐私、治理、AI 监管和执行复杂性;只是它们指标更多、公开程度更高。对 Pathos 而言,实际打破投资逻辑的触发点很直接:与 Tempus 的伙伴关系中断、pocenbrodib 早期临床表现不佳、P-500 或 DO-2 未能变成可信下一波项目,或证据显示内部治理落后于成熟制药伙伴要求。在这些失败模式被私有尽调或公开运营指标证伪之前,剩余风险评分仍会偏高。[CR024, CR025, CR026, CR027, CR028, CR029]
| 风险 | 可监测触发项 | 阈值 / 事件 | 行动含义 |
|---|---|---|---|
| Tempus 集中度 | 合作范围或条款变化 | 关键数据或入组支持丧失 | 在替代能力拿出证据前,下调平台壁垒判断 |
| Pocenbrodib 临床论点 | 早期研究未达到安全性或疗效门槛 | 安全性不可接受,或疗效不足以支撑下一步扩张 | 重估核心项目估值和平台可信度 |
| 治理成熟度 | 没有验证 / 审查控制的私下证据 | 数据室缺少模型治理文件包 | 将法律和执行风险视为可打破投资论点 |
| 资产组合宽度 | Milademetan / P-500 / DO-2 未能推进,或无解释被降优先级 | 可见管线收窄到单一资产 | 大幅提高集中度折价 |
| 融资依赖 | 可信临床证据点出现前就需要新融资 | 透明度未改善,资本需求却上升 | 假设谈判地位更弱、稀释风险更高 |
投资论点破裂标准用于投资人监测,不是管理层预测。
[CR013, CR014, CR024, CR032, CR039, CR040]法律、数据、合作方和融资冲击,都会传导到试验速度、临床证明和未来估值。
[CR002, CR004, CR005, CR013, CR014, CR024]7.5 图表
08估值
8.1 估值背景:Pathos 拿到溢价私募估值,但没有公开财务数据
Pathos 最新公开融资为估值分析设定起点。公司在 2025 年 5 月宣布完成 $365 million Series D,投后估值约 $1.6 billion;就在 7 个月前,它刚宣布以 $600 million 投后估值完成 $62 million Series C。这意味着短期内披露估值约上调 2.7x;对于一家仍不披露收入、ARR、利润率或现金消耗的公司,这是一轮有意义的重新定价。结果是,估值故事由融资动能、战略叙事和平台可选性驱动,而不是传统财务证据。 SEC 记录提供方向性帮助,但不足以填补缺口。Form D 文件确认公司在 2023、2024 和 2025 年反复融资,但没有说明最新一轮中多少是新股、多少是老股交易,普通股之上有哪些优先权或投资者权利,或 Pathos 在进入 Tempus 与 AstraZeneca 合作后还剩多少现金。因此,$1.6 billion 估值是有用信号,但还不是完全可承销价格。投资者应把它视为协商形成的私募估值,仍需用公开可比公司和情景分析三角校准。[CV001, CV002, CV003, CV005, CV006, CV007]
| 论点 | 改变看法的因素 |
|---|---|
| Pathos 已募集大额资本,并拿下可信战略合作方 | 更多公开经济性,以及可重复平台价值证据,会加强正方论点 |
| 公司尚未披露收入,却已经按严肃 AI 生物科技平台给自己定价 | 公开收入、稳固合作关系,或更低价格会缓解这一担忧 |
| Tempus / AstraZeneca 合作支撑了有意义的平台溢价 | 合作伙伴势头减弱,或供应商集中度上升,都会削弱这部分溢价 |
| 融资条款不透明、Series D 参与方未披露,是真实的投资判断缺口 | 披露股权结构表、参与方和优先权,会提高判断把握 |
反向逻辑对价格敏感:同一家公司,如果进入价明显更低,或证据更扎实,就可能更有吸引力。
[CV004, CV021, CV022, CV024, CV025, CV039]建议来自融资强度、缺失的经济指标、公开可比公司比较和情景分散度。
[CV001, CV003, CV010, CV020, CV032, CV036]8.2 公开可比公司显示,Pathos 已按严肃 AI 生物科技平台定价
最干净的公开可比组显示,尽管没有公开收入,Pathos 仍处在上市 AI 生物科技估值阵列的中间。Tempus 在 2026 年 5 月下旬市值约 $8.29 billion,支撑来自 $348.1 million 的 Q1 收入,以及接近 $1.6 billion 的全年收入指引。Caris 报告 Q1 收入 $216.2 million,并在 2025 年 10-K 中披露非关联方持有股份的合计市值约 $3.87 billion。Recursion 约 $1.6 billion,基本等于 Pathos 私募估值;Schrödinger 和 Absci 低于 $1.0 billion,Relay 更接近 $2.9 billion。 这个比较不能证明 Pathos 被高估,但能证明 Series D 要求投资者把公司按不只是早期实验载体来承销。Pathos 的估值更接近上市平台型生物科技同行,而不是典型不透明、收入前私营生物科技公司。因此,可比表把判断推向「偏紧但不离谱」区间:如果平台真能跨多个资产复利,这个估值说得通;但相对于今天的公开运营披露水平,要求很高。[CV011, CV012, CV013, CV014, CV015, CV016]
| 可比公司 | 指标 | 倍数 / 估值 / 状态 | 参考价值 | 局限 |
|---|---|---|---|---|
| Tempus AI | ~$8.29B 市值;Q1 收入 $348.1M | 已规模化的上市 AI 精准医疗平台 | 数据 + AI 肿瘤学规模的最佳公开参照 | 商业成熟度远高于 Pathos |
| Caris Life Sciences | ~$3.87B 总市值;Q1 收入 $216.2M | 上市精准肿瘤诊断同业 | 展示已有收入同业可能呈现的形态 | 业务组合不同,也承担上市公司义务 |
| Recursion Pharmaceuticals | ~$1.6B 市值 | 最接近 Pathos 当前估值标记的上市市值参照 | 可用于 AI 生物科技平台对比 | 有公开管线,也承受市场盯市波动 |
| Schrödinger | ~$0.99B 市值 | AI 赋能平台型生物科技的较低公开基准 | 可作为下行情景锚点 | 模式和软件收入组合不同 |
| Relay Therapeutics | ~$2.90B 市值 | 没有 Tempus 级收入的较高生物科技平台参照 | 显示市场愿意为平台可选性付费 | 不是精确的 AI 数据类比对象 |
| Absci | ~$795M 市值 | 较小的 AI 生物科技基准 | 可作规模化前乐观预期的下限 | 模态和客户模式不同 |
这组可比公司并不完整,但有明确取舍:重点放在带有 AI / 数据 / 精准肿瘤学属性的上市或新近上市公司。
[CV011, CV012, CV013, CV015, CV016, CV017]8.3 结果离散度极大,情景分析比点估值更重要
单一估值数字会掩盖本案核心事实:Pathos 的结果离散度异常大。乐观情景很容易写。如果肿瘤基础模型逻辑跑通,如果 Tempus/AstraZeneca 合作产出可复制洞见,如果 pocenbrodib、P-500 和更新来源的资产展现真实临床杠杆,那么当前私募估值可能显得保守。在那个世界里,Pathos 会从 AI 药物开发叙事升级为多资产、数据护城河平台,足以支撑数十亿美元级上市估值。 悲观情景同样直接。如果早期临床项目令人失望,如果 Pathos/Tempus 关系比预期更昂贵或更不持久,如果公开市场继续压缩 AI 生物科技倍数,或隐藏融资条款稀释未来投资者,那么当前估值可能偏激进。因此,敏感性条和估值区间图采用粗略公开区间,而不是虚假精确的 DCF。在本章框架下,Pathos 看起来是一只值得跟踪的标的:上行真实存在,但当前价格已经假设的成功多于公开记录目前能验证的成功。[CV021, CV022, CV023, CV024, CV025, CV029]
| 情景 | 假设 | 估值 / 回报逻辑 | 关键风险 | 概率信号 |
|---|---|---|---|---|
| 悲观 | 临床证据不及预期,合作伙伴集中度反噬,公开市场倍数压缩 | $0.7B-$1.0B;当前估值标记会显得过高 | 核心资产失手、融资压力悬顶、稀释 | 并非小概率,因为公开经济性缺位 |
| 基准 | Pathos 保住战略合作并推进项目,但证据还不足以支撑大幅扩大溢价 | $1.4B-$1.8B;接近当前估值标记 | 执行滑坡、可比公司倍数走弱 | 最符合当前证据 |
| 乐观 | 平台在多项资产上形成复利,合作伙伴拓展继续放大 | $2.5B-$3.5B;若临床证据验证系统,上行空间打开 | 需要可重复证据,而不只是叙事 | 有可能,但尚未验证 |
情景区间锚定公开可比公司和当前披露,而不是假装精确的 DCF。
[CV024, CV025, CV032, CV033, CV034, CV035]Pathos 缺少宽泛的公开财务基础,因此少数驱动因素主导估值方向。
[CV022, CV024, CV025, CV029, CV032, CV039]当前标记估值接近基准情景;若证明产出滞后,下行空间真实存在,只有平台能跨资产复利时才有显著上行。
[CV032, CV033, CV034, CV035]8.4 建议:观察,中等信心,估值偏高
这里的建议是观察,信心中等,估值立场偏高。这不是否定公司。Pathos 已经融到可观资本,组建了可识别的战略伙伴,并搭起了比许多 AI 生物科技故事更强的平台叙事。但当前公开记录仍然融资证据强于经济性证据,战略雄心强于已验证临床证明。因此,对一家可能变得很有价值、但尚未让当前入场价变得容易承销的公司,观察是合适评级。 什么会推高评级?明确证据显示 pocenbrodib 或后续资产受益于 PathOS 驱动的患者选择;伙伴关系的经济性和持久性更透明;以及可信的数据室材料,覆盖现金、股权结构表、优先权和项目级支出。什么会压低评级?融资环境走弱、Tempus 伙伴关系中断、临床结果失望,或迹象显示当前估值内含公开文件看不到的治理或稀释包袱。在这些变量解决前,入场纪律应保持严格。[CV024, CV025, CV036, CV037, CV038, CV039]
| 建议 | 置信度 | 风险评级 | 估值立场 | 决策含义 |
|---|---|---|---|---|
| 跟踪 | 中 | 高 | 偏高 | 在积极押注前,等待更强证据或更好价格 |
建议反映价格敏感度和公开证据敏感度,而不只反映公司质量。
[CV036, CV037, CV038, CV041]| 触发因素 | 阈值 | 对投资逻辑的传导 | 行动含义 |
|---|---|---|---|
| 核心项目失败 | Pocenbrodib 未能拿出可信证据,或出现重大延迟 | 削弱平台可信度和下一轮融资叙事 | 下调至回避,直到证据或价格重置 |
| 合作伙伴扰动 | Tempus 数据 / 费用关系弱化或发生重大变化 | 伤及核心平台经济性和供应商稳定性 | 立即提高集中度折价 |
| 下轮融资降价或条款不利于内部股东 | 下一轮融资低于当前估值标记,或背上沉重优先权负担 | 说明当前估值过于乐观 | 将预期回报大幅下调 |
| 治理悬而未决 | 数据室披露不利的优先股堆叠,或限制性投资人权利 | 新资金拿到的经济价值低于名义估值 | 暂停投资判断,直到条款重新谈判或完成建模 |
这些触发因素聚焦会在当前价格附近明确改变投资判断的变量。
[CV025, CV039, CV040]| 主题 | 缺失证据 | 重要性 | 负责人 / 尽调路径 |
|---|---|---|---|
| 股权结构表和优先权 | 最新股权结构表、投资人权利、清算优先权、反稀释条款 | 决定名义估值之下的真实经济价值 | 管理层 / 法务数据室 |
| 现金和现金跑道 | 现金余额、月烧钱速度、分情景现金跑道 | 用于判断融资风险和时间压力 | 财务团队 / 董事会材料 |
| 收入和合同 | 任何付费平台、数据或合作伙伴收入,以及合同结构 | 会显著改变估值方法和判断把握 | 财务 / 商业尽调 |
| 项目预算和里程碑 | 按资产拆分的支出和预期证据节点 | 用于把资本投入和价值创造挂钩 | R&D 规划文件 |
| Series D 构成 | 新股 vs 老股交易、参与方名单、战略 vs 财务投资人组合 | 提高对市场信号质量的把握 | 融资文件 / 投资人名单 |
这五项都拿到之前,不能把 Series D 估值标记当成可完全承做的进入价。
[CV039]以当前价格看,Pathos 在战略可选性上得分较好,在经济透明度上得分较弱。
[CV021, CV023, CV024, CV036, CV037, CV038]8.5 图表
免责声明
本报告仅供参考,不构成投资建议。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Pathos AI, Inc. is incorporated in Delaware under its SEC CIK number 0001967854. | 高 | SO012, SO013 |
| CO002 | Pathos AI's principal place of business and mailing address is 600 W. Chicago Ave., Suite 510, Chicago, Illinois 60654, with telephone number 312-765-7820. | 高 | SO012, SO013 |
| CO003 | Pathos AI's IRS EIN is 852945509 and its fiscal year end is December 31. | 高 | SO012, SO013 |
| CO004 | Pathos AI was founded in 2022 by Eric Lefkofsky and Ryan Fukushima. | 中 | SO002 |
| CO005 | Pathos AI was co-founded by Eric Lefkofsky and Ryan Fukushima after they recognized AI's potential impact on drug discovery and development. | 中 | SO002, SO001 |
| CO006 | Eric Lefkofsky is the founder and CEO of Tempus AI, Inc. (Nasdaq: TEM), a health technology company; Pathos was founded by Tempus AI executives, not as a Tempus corporate spinoff. | 中 | SO018, SO002 |
| CO007 | Iker Huerga joined Pathos AI as Chief Executive Officer and Board Member in May 2025, at the time of the Series D announcement. | 高 | SO005, SO017 |
| CO008 | Iker Huerga is a cancer survivor and biotech veteran who most recently served as Chief Data Scientist for Oncology R&D at AstraZeneca prior to joining Pathos AI. | 中 | SO002, SO005 |
| CO009 | Dr. Jens Renstrup serves as Chief Medical Officer of Pathos AI and led the clinical strategy and public statements for the pocenbrodib Phase 1b/2a trial initiation. | 中 | SO009, SO026 |
| CO010 | Pathos AI raised $365 million in a Series D financing round announced May 15, 2025, bringing the post-money valuation to approximately $1.6 billion. | 高 | SO005, SO017, SO018, SO032 |
| CO011 | The Pathos AI Series D was announced on May 15, 2025 with a press release distributed via Globe Newswire. | 高 | SO005, SO017 |
| CO012 | The Pathos AI Form D for the Series D, filed May 1, 2025, shows a total offering of $399,999,933 with $282,999,950 sold to 11 investors as of the filing date. | 中 | SO013 |
| CO013 | Pathos AI raised a $62 million oversubscribed Series C in October 2024, led by New Enterprise Associates (NEA), at a $600 million post-money valuation; Revolution Growth, Lightbank, and Builders VC also participated. | 高 | SO006, SO014, SO017 |
| CO014 | After the Series C close in October 2024, Pathos AI reported total funding of $102 million, implying approximately $40 million raised prior to the Series C. | 中 | SO006 |
| CO015 | Pathos AI filed its first SEC Form D on March 2, 2023, covering the initial fundraise of approximately $40 million (file number 021-474736). | 高 | SO015, SO012 |
| CO016 | Pathos AI's confirmed Series C investors include New Enterprise Associates (NEA), Revolution Growth, Lightbank, and Builders VC. | 中 | SO006 |
| CO017 | Mohamad Makhzoumi, Co-CEO of New Enterprise Associates (NEA), serves as a Pathos AI board member and was quoted in the Tempus/AstraZeneca collaboration announcement. | 中 | SO010, SO025, SO031 |
| CO018 | The PathOS platform is organized around three engines: Scout (AI-enabled asset selection), Sprint (autonomous clinical execution pods), and Foundry (the shared AI core and oncology foundation model). | 中 | SO003, SO001 |
| CO019 | Pathos AI claims access to over 200 petabytes of multimodal oncology data linked to patient outcomes, which the company states is approximately 50× the size of The Cancer Genome Atlas (TCGA). | 低 | SO003 |
| CO020 | Pathos AI states its dataset is approximately 50× the size of The Cancer Genome Atlas (TCGA), the largest public genome dataset in oncology. | 低 | SO003 |
| CO021 | Pocenbrodib (P-300, formerly FT-7051) is an oral, small molecule CBP/P300 inhibitor originally developed by Forma Therapeutics, acquired by Novo Nordisk, and licensed worldwide to Pathos AI. | 中 | SO007, SO017 |
| CO022 | Pathos AI dosed the first patient in its Phase 1b/2a clinical trial of pocenbrodib (NCT06785636) on March 20, 2025, in metastatic castration-resistant prostate cancer (mCRPC). | 高 | SO009, SO023, SO026 |
| CO023 | The pocenbrodib Phase 1b/2a trial (NCT06785636) plans to enroll approximately 203 patients with mCRPC who have received prior anti-androgen therapy. | 中 | SO009, SO026 |
| CO024 | P-500 (PRT811) is a brain-penetrant, selective, oral SAM-competitive PRMT5 inhibitor licensed worldwide from Prelude Therapeutics in August 2024, targeting high-grade glioma and uveal melanoma. | 中 | SO022, SO024 |
| CO025 | Pathos AI completed its acquisition of Rain Oncology (Nasdaq: RAIN) on January 26, 2024, through its subsidiary WK Merger Sub, Inc., at $1.16 per share plus one contingent value right per share. | 中 | SO008 |
| CO026 | As of the tender offer expiration, 28,031,182 shares of Rain Oncology common stock had been tendered, representing approximately 77% of outstanding Rain shares. | 中 | SO008 |
| CO027 | Rain Oncology's primary clinical asset, milademetan, is a small molecule, oral inhibitor of the p53-MDM2 complex that had reached Phase 3 development at the time of the Rain acquisition, but has not been publicly referenced in Pathos communications since the acquisition closed. | 中 | SO008, SO017 |
| CO028 | On April 23, 2025, Pathos AI announced multi-year strategic collaborations with AstraZeneca and Tempus AI to build a multimodal foundation model in oncology that will be shared among all three parties. | 高 | SO010, SO025, SO031 |
| CO029 | The Pathos–AstraZeneca–Tempus collaboration agreements include $200 million in data licensing and model development fees payable to Tempus AI. | 高 | SO010, SO025, SO031 |
| CO030 | On May 6, 2026, Pathos AI announced the acquisition of a majority stake in DeuterOncology, a Belgium-based company developing DO-2, a deuterated third-generation MET kinase inhibitor. | 中 | SO011, SO021 |
| CO031 | In a Phase 1 study, DO-2 demonstrated 100% tumor shrinkage in all 10 evaluable MET exon 14 skipping NSCLC patients, with zero Grade 4 adverse events and a peripheral edema rate of 5%. | 低 | SO021, SO011 |
| CO032 | DO-2 was identified, evaluated, and brought to acquisition recommendation entirely through the Pathos Foundry AI platform after being flagged as a top-ranked candidate in late 2025. | 低 | SO011, SO021 |
| CO033 | Pathos AI was founded by executives from Tempus AI and is a legally distinct, independently operated company; it is not a corporate spinoff or subsidiary of Tempus AI. | 中 | SO018, SO002 |
| CO034 | GenomeWeb published a correction to an earlier version of its Series D article, clarifying that Pathos AI was founded by Tempus AI executives, not as a Tempus AI spinoff. | 中 | SO018 |
| CO035 | Pathos AI describes itself as a clinical-stage AI and technology company advancing its own pipeline of cancer therapies through a combination of AI-guided asset selection and clinical execution. | 中 | SO001, SO011 |
| CO036 | Ryan Fukushima served as Pathos AI's founding CEO and interim CEO; his exact role following Iker Huerga's appointment in May 2025 has not been publicly disclosed. | 中 | SO006, SO005 |
| CO037 | Pathos AI's SEC-registered phone number is 312-765-7820 and its fiscal year end is December 31. | 高 | SO012, SO013 |
| CO038 | Pathos AI states that the Foundry engine's backbone is the oncology foundation model being built in partnership with Tempus and AstraZeneca, described as the largest oncology foundation model in the industry. | 低 | SO003, SO010 |
| CO039 | DeuterOncology is a Belgium-based company and its DO-2 candidate has completed Phase 1 dose escalation across eight clinical sites in the Netherlands, Belgium, and France. | 中 | SO011, SO021 |
| CO040 | Pathos AI's official website is www.pathos.com and the company's business development contact is bd@pathos.com. | 中 | SO001 |
| CM001 | IQVIA's Global Oncology Trends 2025 reports global oncology medicine spending reached $252B in 2024 and is expected to reach $441B by 2029, growing at approximately 11.9% CAGR driven by novel modalities including ADCs and bispecific antibodies. | 中 | SM001 |
| CM002 | MarketsandMarkets estimates the AI in oncology market at $2.45B in 2024, projected to reach $11.52B by 2030 at a 29.4% CAGR, encompassing drug discovery, diagnostics, clinical decision support, and imaging AI. | 中 | SM002 |
| CM003 | Mordor Intelligence sizes the real-world evidence solutions market at $2.44B in 2025, projected to reach $6.04B by 2031 at a 16.33% CAGR, with cloud deployment capturing 64.35% share and North America leading at 40.95% regional share. | 中 | SM003 |
| CM004 | MarketsandMarkets estimates the drug discovery technologies market at $30.58B in 2025, projected to reach $51.51B by 2030 at 11.0% CAGR, driven by advanced screening platforms and rising demand for biologics and RNA-based drugs. | 中 | SM002 |
| CM005 | MarketsandMarkets estimates the global precision diagnostics and medicine market at $145.53B in 2024, projected at $246.66B by 2029 at 11.1% CAGR, driven by AI/ML integration and pharma-diagnostics collaborations. | 中 | SM002 |
| CM006 | Allied Market Research estimates the global NGS market at $12.98B in 2023, projected to reach $97.81B by 2035 at 18.3% CAGR, driven by falling sequencing costs and rising clinical utility in oncology. | 中 | SM015 |
| CM007 | American Cancer Society estimates approximately 2,041,910 new cancer cases and 618,120 cancer deaths in the United States in 2025, confirming oncology remains a very large clinical-volume market. | 中 | SM004 |
| CM008 | WHO reports approximately 10 million cancer deaths globally in 2022; lung cancer was the leading cause with 2.5 million new cases, followed by breast (2.3M), colon and rectum (1.9M), and prostate (1.5M). | 中 | SM005 |
| CM009 | CDC's U.S. Cancer Statistics now covers 100% of the U.S. population and recorded 36.7 million new cancer cases from 2003 to 2022, providing a nationwide longitudinal data backbone increasingly relevant to oncology AI model development. | 中 | SM006 |
| CM010 | Mordor Intelligence reports oncology represents 34.65% of the global RWE solutions market by therapeutic area in 2025, and pharmaceutical and medical device companies hold 49.2% of end-user market share. | 中 | SM003 |
| CM011 | IQVIA reports 2,162 oncology trial starts in 2024, up 12% from 2019; 74% of trial starts targeted rare cancers; oncology represents 41% of all global clinical trial starts. | 中 | SM001 |
| CM012 | IQVIA reports 25 oncology novel active substances (NAS) launched globally in 2024, averaging 26 per year from 2020-2024 versus 16 per year in 2015-2019, indicating sustained pipeline productivity and continued demand for AI optimization tools. | 中 | SM001 |
| CM013 | Caris Life Sciences' 10-K for FY2025 discloses that as of December 31, 2025 Caris had surpassed 1 million total molecular profiles, with gross margin improving from 47% in Q1 2025 to 65% in Q1 2026, demonstrating rapid scale and margin expansion. | 高 | SM008, SM009 |
| CM014 | Caris Life Sciences reported Q1 2026 total revenue of $216.2M, up 79% year-over-year, with approximately 52,800 clinical profiling cases completed in the quarter and adjusted EBITDA of $26.2M positive. | 高 | SM009, SM008 |
| CM015 | Tempus AI reported Q1 2026 revenue of $348.1M (+36.1% YoY), with data and applications revenue of $87M (+40.5% YoY); the company guided to $1.59-1.60B in FY2026 revenue representing ~25% annual growth. | 中 | SM007 |
| CM016 | Tempus AI signed multi-year strategic collaborations with Merck and Gilead in Q1 2026 for enterprise-wide access to multimodal oncology data and AI analytics, demonstrating active Big Pharma demand for oncology AI platform services. | 中 | SM007 |
| CM017 | ConcertAI raised $150M in June 2024 at an approximately $1.9B valuation, led by Goldman Sachs Asset Management, underscoring ongoing private investment in oncology RWE platforms. | 中 | SM012 |
| CM018 | Roche Group acquired Flatiron Health for approximately $1.9B in 2018, establishing a high-value precedent for biopharma acquisition of oncology real-world data platforms and embedding RWD in Roche's R&D strategy. | 中 | SM017 |
| CM019 | Pathos AI claims a dataset of 200+ petabytes of multimodal oncology data, described as approximately 50× the scale of TCGA, encompassing genomic, imaging, and clinical records; this figure appears exclusively in company-authored materials with no independent verification. | 低 | SM014, SM019 |
| CM020 | A 2025 Springer systematic review of AI in oncology found significant barriers to clinical adoption including algorithmic bias, GDPR/HIPAA compliance requirements, unclear liability frameworks for AI-driven recommendations, and a lack of prospective multi-centre validation. | 中 | SM013 |
| CM021 | Pathos AI operates at the intersection of AI-assisted oncology drug discovery, oncology RWE, and precision oncology diagnostics; the company was pre-commercial in therapeutics as of the May 2025 Series D, with no disclosed platform product revenues. | 中 | SM014, SM018 |
| CM022 | Foundation Medicine, backed by Roche, markets comprehensive genomic profiling solutions for cancer care and remains a primary incumbent in the oncology CGP segment where Caris, Tempus, and specialty diagnostics companies compete. | 中 | SM024 |
| CM023 | Tempus AI and Caris Life Sciences together represent the two largest pure-play oncology AI/data companies by disclosed revenue; combined Q1 2026 revenues of approximately $564M imply an annualized commercial oncology data services market of approximately $2.3B for the two companies alone. | 中 | SM007, SM009 |
| CM024 | Large pharmaceutical companies allocate AI and data platform spending from R&D budgets managed by Chief Digital Officers or SVP R&D; multi-year enterprise agreements with annual data access fees are the demonstrated commercial model, with Tempus AI disclosing multi-year agreements with Merck, Gilead, AstraZeneca, and others. | 中 | SM007, SM010 |
| CM025 | Pathos AI, AstraZeneca, and Tempus AI announced strategic agreements in April 2025 to develop what the parties describe as the largest multimodal foundation model in oncology, with AstraZeneca and Pathos AI collectively paying up to $200M to Tempus AI for data and modeling services. | 中 | SM018, SM020, SM021 |
| CM026 | The FDA's Real-World Evidence Framework (2018) and subsequent regulatory guidance formally accepted RWE as an evidentiary standard for certain drug approval and label expansion submissions, creating regulatory tailwind for oncology data platform companies. | 中 | SM025, SM013 |
| CM027 | The NGS sequencing cost trajectory has enabled large-scale multi-omic datasets at declining per-sample cost; the NGS market is projected to grow from $12.98B in 2023 to $97.81B by 2035 at 18.3% CAGR, driven by both clinical and research oncology adoption. | 中 | SM015 |
| CM028 | Oncology drug failure rates remain very high; academic literature and the Springer (2025) review reference approximately 5% success rate from Phase 1 to regulatory approval for oncology drugs, creating strong commercial demand for AI-assisted asset selection and trial optimization platforms. | 中 | SM013 |
| CM029 | HIPAA in the United States and GDPR in Europe constrain cross-institutional patient data sharing for AI training, requiring de-identification frameworks and, in federated architectures such as Owkin's, privacy-preserving compute that does not centralize raw patient data. | 中 | SM013, SM027 |
| CM030 | High switching costs characterize the oncology RWD market; biopharma partners embed multi-omic longitudinal patient data into multi-year research programs, creating platform stickiness for Tempus AI and Caris Life Sciences that new entrants must overcome with demonstrably superior data coverage or algorithms. | 中 | SM007, SM008 |
| CM031 | Grand View Research forecasts significant growth in the AI in clinical trials market through 2030, driven by demand for faster patient enrollment via AI screening, synthetic control arms, and biomarker-driven eligibility criteria; a specific market size value was not available in the fetched content. | 低 | SM016 |
| CM032 | The Springer (2025) systematic review notes that AI clinical decision-support evaluation in oncology is predominantly conducted on retrospective datasets from single-centre studies, limiting generalizability and real-world clinical utility claims. | 中 | SM013 |
| CM033 | Syapse operates a real-world oncology outcomes data network serving community health systems; Owkin builds federated AI solutions for oncology biopharma; ConcertAI aggregates oncology real-world patient data for pharma R&D — illustrating the diversity of vendor models and buyer channels in the oncology RWE market. | 中 | SM023, SM027, SM011 |
| CM034 | IQVIA projects oncology medicine spending growth will slow after 2027 as biosimilar competition for PD-1/PD-L1 backbone therapies begins, potentially moderating total oncology R&D investment growth rates in 2028-2029. | 中 | SM001 |
| CM035 | No publicly available independent third-party verification of Pathos AI's dataset quality, de-identification standards, data completeness, or claimed 200+ petabyte scale has been identified in the research corpus; the figure relies solely on company press releases. | 低 | |
| CM036 | Pathos AI's estimated serviceable obtainable market (SOM) in AI-assisted oncology drug discovery is below $200M at current stage, based on industry analogies for early-commercial AI drug discovery platforms; at the Series D post-money valuation of approximately $1.6B, the company trades at a premium to SOM reflecting pipeline asset and platform optionality. | 低 | SM014, SM022 |
| CM037 | Analyst market size estimates for oncology AI services span more than 50× depending on boundary definition — from the oncology RWE slice (~$846M per Mordor) to the full global oncology drug economy ($252B per IQVIA) — illustrating that any single TAM figure is meaningless without a stated boundary definition. | 中 | SM001, SM003 |
| CM038 | Large pharmaceutical companies including AstraZeneca, Merck, and Gilead have signed enterprise multi-year oncology AI platform agreements with Tempus AI, demonstrating that Big Pharma willingness to pay at scale for oncology AI data services is real and commercially active as of 2025-2026. | 中 | SM007, SM010 |
| CM039 | Tempus AI's commercial trajectory — from single data-access pilots to enterprise platform licensing to foundation model co-development — illustrates a multi-year adoption funnel for oncology AI data platforms, implying Pathos AI faces a 2-4 year ramp to meaningful pharma contract revenues even with a differentiated dataset. | 低 | SM007, SM020 |
| CP001 | Pathos says it is redefining drug development starting in oncology. | 中 | SP001 |
| CP002 | Pathos publicly presents a precision oncology pipeline rather than a public software price list. | 中 | SP002 |
| CP003 | Recursion and Exscientia announced a definitive agreement in August 2024 to combine into a global technology-enabled drug discovery leader. | 高 | SP003, SP004 |
| CP004 | By November 2024 Nasdaq coverage described Recursion and Exscientia as officially combined. | 中 | SP004 |
| CP005 | The Recursion-Exscientia combination is positioned around end-to-end discovery capabilities rather than clinical asset rehabilitation. | 中 | SP003, SP004 |
| CP006 | Insilico’s pipeline page lists more than 40 total programs. | 中 | SP005 |
| CP007 | Insilico’s pipeline page says 13 pipelines have received IND approval. | 中 | SP005 |
| CP008 | Insilico’s Lilly collaboration grants an exclusive worldwide license for a portfolio of programs. | 中 | SP006 |
| CP009 | Insilico says the Lilly collaboration can reach approximately $2.75 billion plus tiered royalties. | 中 | SP006 |
| CP010 | BenevolentAI described itself in April 2024 as a leader in applying advanced AI to accelerate biopharma drug discovery. | 中 | SP007 |
| CP011 | BenevolentAI’s Merck collaboration covers three targets in oncology, neurology, and immunology. | 中 | SP008 |
| CP012 | BenevolentAI announced a major strategic overhaul in December 2024 to return to its founding TechBio mission. | 中 | SP009 |
| CP013 | BenevolentAI proposed delisting via merger with Osaka Holdings in 2025. | 中 | SP010 |
| CP014 | Schrödinger says it offers an industry-leading computational platform for molecular design. | 中 | SP011 |
| CP015 | Schrödinger’s pipeline page lists SGR-1505 as a Phase 1 MALT1 inhibitor program. | 中 | SP012 |
| CP016 | Schrödinger’s 2024 Novartis agreement includes $150 million upfront plus up to about $2.3 billion in milestones and royalties. | 中 | SP013 |
| CP017 | Relay’s pipeline page features zovegalisib (RLY-2608) and lirafugratinib (RLY-4008). | 中 | SP014 |
| CP018 | Relay markets a motion-based discovery engine called the Dynamo platform. | 中 | SP015 |
| CP019 | Elevar’s 2024 announcement says it licensed lirafugratinib from Relay under a global agreement. | 中 | SP016 |
| CP020 | Valo describes its approach as AI-enabled human causal biology and closed-loop chemistry. | 中 | SP017 |
| CP021 | Business Wire said Valo recruited former Novo Nordisk R&D leader Karin Conde-Knape as CSO in 2026. | 中 | SP018 |
| CP022 | Ikena and Inmagene announced a merger and private placement in December 2024. | 高 | SP019, SP020 |
| CP023 | After the merger closed in July 2025, the combined company adopted the ImageneBio name and IMA ticker with $75 million of concurrent financing. | 高 | SP019, SP020 |
| CP024 | Boundless says ecDNA is a root cause of oncogene amplifications in over 17% of cancer patients. | 中 | SP021 |
| CP025 | Boundless describes itself as a clinical-stage precision oncology company. | 中 | SP021 |
| CP026 | Absci’s pipeline includes ABS-201 in Phase 1/2a. | 中 | SP022 |
| CP027 | Tempus integrated molecular profiling directly into Flatiron’s OncoEMR. | 中 | SP023 |
| CP028 | TechCrunch covered Tempus in 2024 as an AI health-tech company heading toward the public markets under Eric Lefkofsky. | 中 | SP024 |
| CP029 | Caris and Flatiron said their joint offering combines Caris DNA, RNA, and imaging data with Flatiron real-world data. | 中 | SP025 |
| CP030 | ConcertAI and Caris said AbbVie would use their combined clinical and genomic databases with AI and machine learning insights to accelerate oncology R&D. | 中 | SP026 |
| CP031 | ConcertAI’s 2025 Precision Suite announcement says it uses CARAai and oncology data to deliver enterprise-wide value to life sciences customers. | 中 | SP027 |
| CP032 | Flatiron highlighted more than 25 research acceptances and next-generation capabilities at ASCO 2026. | 中 | SP028 |
| CP033 | Roche says Foundation Medicine’s comprehensive genomic profiling tests analyze more than 300 genes. | 中 | SP029 |
| CP034 | Roche says Foundation Medicine became an independent Roche affiliate in 2018. | 中 | SP029 |
| CP035 | Owkin’s AstraZeneca partnership described Owkin as an end-to-end AI-biotech using causal AI across drug discovery, development, and diagnostics. | 中 | SP030 |
| CP036 | Owkin expanded the MOSAIC study with 10x spatial-omics and single-cell technologies. | 中 | SP031 |
| CP037 | Sanofi publicly highlighted Owkin as a precision-medicine AI partner. | 中 | SP032 |
| CP038 | Pathos is closer to AI-assisted clinical asset triage than to de novo molecule design. | 中 | SP001, SP002, SP003, SP005, SP014 |
| CP039 | Recursion and Insilico compete with Pathos on AI narrative depth, but they attack earlier parts of the value chain. | 中 | SP003, SP004, SP005, SP006, SP001, SP002 |
| CP040 | Schrödinger and Relay bring stronger molecule-optimization depth than Pathos but do not mirror Pathos’s asset-rehabilitation model. | 中 | SP011, SP012, SP014, SP015, SP001, SP002 |
| CP041 | Tempus, Flatiron, Caris, ConcertAI, and Foundation Medicine compete with Pathos through data and workflow distribution rather than direct drug-asset origination. | 中 | SP023, SP025, SP026, SP028, SP029 |
| CP042 | Owkin and Valo are the closest adjacent analogs because each combines AI platform claims with pharma-facing partnerships, but both are broader than Pathos’s current oncology wedge. | 中 | SP017, SP018, SP030, SP031, SP032, SP001, SP002 |
| CP043 | Ikena’s disappearance into ImageneBio and BenevolentAI’s strategic reset show that AI-biotech threat levels can shrink quickly when proof and capital diverge. | 中 | SP009, SP010, SP019, SP020 |
| CP044 | Public pricing across the peer set is mostly expressed as collaboration economics, milestone ranges, or enterprise integrations instead of standard rate cards. | 中 | SP006, SP013, SP016, SP023, SP026, SP027 |
| CP045 | Pathos currently lacks the workflow lock-in that Tempus and Flatiron have and the software monetization transparency that Schrödinger has. | 中 | SP013, SP023, SP028, SP001, SP002 |
| CP046 | Pathos’s strongest defendable differentiation is the combination of asset acquisition, biomarker-led development, and oncology-specific execution. | 中 | SP001, SP002, SP003, SP014 |
| CP047 | Pathos’s moat is not yet durably proven because public evidence remains heavier on partnerships, acquisitions, and platform descriptions than on completed Pathos outcome cycles. | 中 | SP001, SP002, SP023, SP027 |
| CP048 | If Pathos can show repeated improvement in trial design or patient-selection outcomes, it could occupy a distinct middle ground between discovery platforms and data incumbents. | 低 | SP001, SP002, SP023, SP028 |
| CI001 | Pathos announced a $365 million Series D at an approximately $1.6 billion post-money valuation. | 高 | SI001, SI006 |
| CI002 | The 2025 Form D recorded a first sale date of 2025-04-17, a $399,999,933 total offering amount, $282,999,950 sold, and 11 investors. | 高 | SI005, SI006 |
| CI003 | Pathos announced a $62 million Series C led by NEA at a $600 million post-money valuation. | 高 | SI002, SI007 |
| CI004 | The 2024 Form D recorded a first sale date of 2024-10-24, a $61,999,979 total offering amount, and 13 investors. | 高 | SI005, SI007 |
| CI005 | The 2023 Form D recorded a first sale date of 2023-02-17, a $39,999,988 total offering amount, $19,999,992 sold at filing, and 4 investors. | 高 | SI005, SI008 |
| CI006 | Pathos said total funding reached $102 million after the Series C, implying about $40 million of pre-Series-C capital. | 中 | SI002, SI008 |
| CI007 | The SEC submissions record shows three Form D notices under CIK 0001967854 through the run date. | 中 | SI005 |
| CI008 | No debt facility, credit line, or project-finance obligation surfaced in the reviewed official and filing sources. | 中 | SI001, SI002, SI005 |
| CI009 | Pathos disclosed $200 million of data-licensing and model-development fees payable to Tempus. | 中 | SI004 |
| CI010 | Pathos said Series D proceeds would support the clinical-stage pipeline and further investment in its oncology foundation model. | 中 | SI001 |
| CI011 | The Rain Oncology acquisition shows Pathos has used M&A as a capital-deployment tool to add pipeline assets. | 中 | SI009 |
| CI012 | The DeuterOncology acquisition shows Pathos continued deploying capital into AI-sourced asset acquisition in 2026. | 中 | SI029 |
| CI013 | Pathos said DO-2 was one of four major portfolio decisions made through Foundry in Q1 2026. | 中 | SI029 |
| CI014 | No product revenue, ARR, or revenue mix has been publicly disclosed by Pathos. | 中 | SI001, SI002, SI004, SI005 |
| CI015 | Pathos has not published price points or contract economics for platform access or AI trial-design services. | 中 | SI001, SI002, SI004 |
| CI016 | The clearest publicly disclosed collaboration economics show Pathos paying fees rather than receiving them. | 中 | SI004 |
| CI017 | Public GTM appears centered on biopharma partnering and asset acquisition rather than scaled software sales. | 中 | SI003, SI004, SI029 |
| CI018 | No public CAC, payback, sales-cycle, renewal-rate, or channel-economics metrics have been disclosed. | 中 | SI001, SI002, SI004 |
| CI019 | Public traction for Pathos is fundraising and portfolio expansion rather than disclosed commercial revenue metrics. | 中 | SI001, SI002, SI009, SI029 |
| CI020 | Any Pathos monetization thesis depends on future licensing, partnering, milestones, royalties, or asset exits rather than disclosed recurring revenue. | 中 | SI001, SI004, SI015 |
| CI021 | Public sources do not disclose units, locations, active users, or utilization metrics that could substitute for revenue traction. | 中 | SI001, SI002, SI004, SI005 |
| CI022 | Pathos should be modeled as a clinical-stage biotech cost structure rather than as a low-cost software platform. | 中 | SI001, SI004, SI020, SI029 |
| CI023 | Clinical-trial cost literature says delayed activation, poor accrual, and infrastructure burden can materially waste budgets. | 中 | SI025, SI026 |
| CI024 | Public oncology accrual analysis shows enrollment speed is a meaningful driver of trial duration and financing need. | 中 | SI027 |
| CI025 | JAMA Network Open estimated mean drug-development cost at $172.7 million before higher failed-program and capital-cost scenarios are included. | 中 | SI028 |
| CI026 | Relay reported approximately $710 million of cash, cash equivalents, and investments at the end of Q1 2025. | 中 | SI021 |
| CI027 | Relay said its Q1 2025 cash position extended runway into 2029. | 中 | SI021 |
| CI028 | Relay reported $642.1 million of cash, cash equivalents, and investments at the end of Q1 2026. | 中 | SI019 |
| CI029 | Relay disclosed $137.1 million of Q1 2026 ATM proceeds, underscoring ongoing capital-markets dependence. | 中 | SI019 |
| CI030 | Schrödinger reported $406 million of cash, cash equivalents, restricted cash, and marketable securities at the end of Q1 2026. | 中 | SI017 |
| CI031 | Schrödinger reported a $60.0 million net loss in Q1 2026. | 中 | SI017 |
| CI032 | Insilico reported cash and bank balances of $393.3 million as of December 31, 2025. | 中 | SI022 |
| CI033 | Insilico reported 2025 revenue of $56.2 million. | 中 | SI022 |
| CI034 | A reasonable public benchmark for Pathos burn is roughly $120 million to $240 million annually. | 低 | SI017, SI019, SI021, SI022, SI025, SI028 |
| CI035 | On that benchmark, the $365 million Series D alone could support roughly 18 to 36 months of burn before considering opening cash or obligation timing. | 低 | SI001, SI004, SI017, SI019, SI021, SI022, SI025, SI028 |
| CI036 | Pathos has not disclosed gross margin, cost per asset, or working-capital needs. | 中 | SI001, SI002, SI004, SI005 |
| CI037 | Pathos has raised roughly $467 million of cumulative disclosed capital across the 2023, 2024, and 2025 financings. | 高 | SI001, SI002, SI006, SI007, SI008 |
| CI038 | Relative to a roughly $1.6 billion post-money valuation, cumulative disclosed capital equals about 29 percent of Pathos' post-money value. | 低 | SI001, SI002 |
| CI039 | Using four public portfolio decisions or assets as the denominator, Pathos has raised roughly $117 million per disclosed asset. | 中 | SI001, SI002, SI009, SI029 |
| CI040 | Variational AI's $5.5 million seed extension shows that the long tail of AI-discovery financing sits far below Pathos' scale. | 中 | SI010 |
| CI041 | Isomorphic Labs' $2.1 billion Series B shows that Pathos is large but still below the very top tier of AI-drug-discovery financing. | 中 | SI012 |
| CI042 | Insilico's $110 million Series E and 40-plus publicly disclosed programs show a peer that pairs financing with broader public pipeline disclosure than Pathos. | 中 | SI011, SI024 |
| CI043 | Schrödinger's official platform model combines recurring software revenue with drug-discovery upside. | 高 | SI017, SI018 |
| CI044 | Pathos looks structurally closer to Relay's capital-intensive clinical-stage oncology model than to Schrödinger's hybrid software model. | 中 | SI017, SI019, SI020, SI021 |
| CI045 | McKinsey notes that first-time biotech launchers increasingly have to build commercialization capability themselves. | 中 | SI013 |
| CI046 | Nature warned that biotech financing was only slowly recovering after a difficult three-year period. | 中 | SI014 |
| CI047 | The Series D announcement said the round included a mix of new and existing investors but did not name them. | 中 | SI001 |
| CI048 | The Tempus and AstraZeneca partnership disclosures do not specify revenue sharing, royalty splits, or economic ownership of future model outputs. | 中 | SI004 |
| CI049 | Public sources do not disclose cash on hand at the Series D close. | 中 | SI001, SI006 |
| CI050 | Public sources do not disclose post-Series-D cap-table terms or liquidation preferences. | 中 | SI001, SI005 |
| CI051 | Public sources do not disclose ownership concentration or named positions for most Series D investors. | 中 | SI001 |
| CI052 | Public sources do not disclose recognized revenue or inbound economics from Rain, milademetan, or DO-2. | 中 | SI009, SI029 |
| CI053 | Pathos is well funded but still effectively pre-revenue, capital intensive, and dependent on private disclosures to prove revenue quality and true runway. | 中 | SI001, SI004, SI014, SI017, SI019, SI021, SI022 |
| CE001 | PathOS is built around three named engines: Foundry, Scout, and Sprint. | 中 | SE001 |
| CE002 | Pathos says the PathOS platform has access to more than 200 petabytes of multimodal oncology data linked to patient outcomes. | 高 | SE001, SE013 |
| CE003 | Scout is described as the asset-selection engine that ranks therapies and patient subgroups using multimodal evidence. | 中 | SE001 |
| CE004 | Sprint is described as the clinical execution engine that moves selected assets from one development milestone to the next. | 中 | SE001 |
| CE005 | Foundry is described as the shared AI core that powers Scout and Sprint and connects to a lab-in-the-loop validation system. | 高 | SE001, SE002 |
| CE006 | Pathos positions the platform as a continuously learning system in which each program teaches the next. | 中 | SE001, SE004 |
| CE007 | The company homepage says Pathos is a clinical-stage biotech using multimodal data, biological modeling, and biopharma partnerships to improve oncology drug development. | 中 | SE002 |
| CE009 | The about page says Iker Huerga joined as CEO in 2025 after leadership roles at AstraZeneca and Tempus. | 中 | SE003 |
| CE010 | Pathos says it uses multimodal data to identify responsive patient subgroups, design adaptive trials, and run them with AI-enabled teams. | 中 | SE003, SE002 |
| CE011 | The collaboration announced in April 2025 is a multi-year strategic agreement among Pathos, Tempus, and AstraZeneca. | 高 | SE006, SE007 |
| CE012 | The collaboration is aimed at building a multimodal foundation model in oncology to support biological insight, target discovery, and therapeutics development. | 高 | SE006, SE007 |
| CE013 | Pathos said Tempus will provide de-identified oncology data for the foundation model and that the finished model will be shared among the three parties. | 高 | SE006, SE007 |
| CE014 | The Pathos and Tempus announcements both say the agreements include $200 million in data-licensing and model-development fees to Tempus. | 高 | SE006, SE007 |
| CE015 | Tempus markets its life-sciences platform as having more than 8.5 million de-identified research records. | 中 | SE008 |
| CE016 | Tempus markets its life-sciences platform as having more than 450 petabytes of unique data elements per patient on average and connections to more than 5,000 healthcare institutions. | 中 | SE008 |
| CE017 | Pathos launched its first clinical-stage asset by licensing FT-7051 from Novo Nordisk and renaming it P-300. | 中 | SE009 |
| CE018 | The same Pathos licensing announcement said FT-7051 was already in phase 1 development when Pathos acquired it. | 中 | SE009 |
| CE019 | Pathos public materials now call the FT-7051 program pocenbrodib. | 中 | SE004, SE009, SE010 |
| CE020 | Pathos said the first patient in its phase 1b/2a pocenbrodib trial was dosed in March 2025. | 高 | SE010, SE011, SE025 |
| CE021 | Pathos said the pocenbrodib trial is designed to enroll about 203 patients with metastatic castration-resistant prostate cancer. | 高 | SE010, SE011 |
| CE022 | The urology summary says the pocenbrodib study is running across three U.S. clinical trial sites. | 中 | SE011 |
| CE023 | The pocenbrodib study tests monotherapy and combinations with abiraterone acetate, olaparib, and 177Lu-PSMA-617. | 高 | SE010, SE011 |
| CE024 | Pathos describes pocenbrodib as its first clinical-stage asset and frames biomarker selection as central to the program strategy. | 中 | SE010 |
| CE025 | Pathos says it used the platform to identify biological mechanisms relevant to P-500 and to map them to outcomes in a subgroup of IDH-positive high-grade glioma patients. | 中 | SE012 |
| CE026 | The Series C announcement described P-500 as a phase-II-ready, brain-penetrant PRMT5 inhibitor. | 中 | SE012 |
| CE027 | The Pathos pipeline page publicly lists only two development programs: pocenbrodib and P-500. | 中 | SE004 |
| CE028 | The Biospace PRT811 report says Pathos obtained a worldwide license to the PRMT5 inhibitor in August 2024 and renamed it P-500. | 中 | SE020 |
| CE029 | The Biospace PRT811 report says two confirmed complete responses were observed among 16 IDH-positive high-grade glioma patients in the prior phase 1 trial. | 中 | SE020 |
| CE030 | The same report says one of the complete responses was ongoing at 31 months and another lasted 7.5 months. | 中 | SE020 |
| CE031 | The same report says one confirmed and one unconfirmed partial response were seen among uveal melanoma patients with SF3B1 mutations. | 中 | SE020 |
| CE032 | The same report says the most common grade 3 or higher adverse events in the phase 1 P-500 safety population were thrombocytopenia, anemia, and fatigue. | 中 | SE020 |
| CE033 | Fierce Biotech and Pharmaphorum both reported that launching the next P-500 trial remained a 2025 priority after the Series D financing. | 高 | SE014, SE016 |
| CE034 | Pathos announced in May 2026 that it acquired a majority stake in DeuterOncology to add the MET inhibitor DO-2 to the pipeline. | 高 | SE017, SE018 |
| CE035 | Pathos said DO-2 was identified and advanced to acquisition through Foundry. | 中 | SE017 |
| CE036 | Pathos said DO-2 showed 100% tumor shrinkage in all evaluable MET exon 14 skipping NSCLC patients in a 28-patient phase 1 study. | 高 | SE017, SE018 |
| CE037 | Pathos said DO-2 had a 5% peripheral edema rate versus 62% to 82% for approved competitors and carries patent exclusivity to December 2040. | 高 | SE017, SE018 |
| CE038 | Pathos said DO-2 is one of four major portfolio decisions made through Foundry in the first quarter of 2026. | 中 | SE017 |
| CE039 | Pathos completed its Rain Oncology acquisition in January 2024, making Rain a wholly owned subsidiary and taking control of milademetan. | 中 | SE019 |
| CE040 | Independent coverage in 2025 continued to describe Pathos as building an oncology foundation model from clinical, molecular, and imaging data rather than as a traditional single-asset biotech. | 高 | SE014, SE015, SE016 |
| CE041 | Pathos has public recruiting infrastructure but no public developer documentation, code repository, or API surface was surfaced in the reviewed official materials. | 中 | SE002, SE003, SE004, SE005 |
| CE042 | HHS guidance says properly de-identified data still retains some residual re-identification risk, which matters for any oncology foundation model built on patient-linked records. | 中 | SE021 |
| CE043 | NCI says biomarker testing is important to precision medicine but not yet part of routine care for most patients, which constrains how quickly biomarker-led trial designs can diffuse. | 中 | SE022 |
| CE044 | The ASCO Post article says legal standards for AI-related diagnostic and treatment errors in oncology remain unsettled. | 中 | SE023 |
| CE045 | The Springer review says AI models in oncology are only as effective as the data used to train them and highlights data privacy, consent, and bias risks. | 中 | SE024 |
| CE046 | The Tempus TIME case study says six of the first ten patients enrolled in Pathos’ pocenbrodib trial were identified through the TIME network. | 中 | SE025 |
| CE047 | Tempus reported that its data-and-applications revenue reached $87.0 million in Q1 2026, suggesting the Pathos collaboration depends on an increasingly commercialized external data platform. | 中 | SE026 |
| CU001 | Pathos does not publicly disclose customer count, ARR, or recurring software revenue on its official site or financing releases. | 中 | SU001, SU002, SU006, SU007 |
| CU002 | Pathos publicly describes itself as partnering with pharma, biotech, academia, and investors rather than as a scaled diagnostics or software vendor with disclosed customer metrics. | 中 | SU002 |
| CU003 | The public Pathos story is built around strategic relationships and sponsored trials, not a broad installed-base narrative. | 中 | SU001, SU002, SU003, SU007 |
| CU004 | The April 2025 Tempus and AstraZeneca collaboration is a named, multi-year external demand signal for Pathos’ platform ambitions. | 高 | SU004, SU009 |
| CU005 | Both Pathos and Tempus say the collaboration includes $200 million in data licensing and model-development fees to Tempus. | 高 | SU004, SU009 |
| CU006 | The same collaboration indicates the resulting oncology foundation model will be shared among Pathos, Tempus, and AstraZeneca. | 高 | SU004, SU009 |
| CU007 | The Tempus TIME case study positions Pathos as a life-sciences sponsor using Tempus infrastructure to accelerate an early-phase oncology trial. | 中 | SU008 |
| CU008 | The TIME case study says Tempus helped identify six of the first ten patient matches on the Pathos pocenbrodib study. | 高 | SU008, SU026 |
| CU009 | Tempus says the TIME network can reduce just-in-time trial activation to as few as about 10 business days. | 中 | SU008 |
| CU010 | Tempus quotes Pathos CEO Iker Huerga saying the network helped identify sites with high volumes of potentially eligible patients and enabled rapid activation. | 中 | SU008 |
| CU011 | Pathos’ pocenbrodib trial is running across three U.S. clinical sites, which shows real but still limited public deployment evidence. | 高 | SU005, SU026 |
| CU012 | Pathos has public evidence of an active sponsored study, but no public evidence of a broad provider or health-system customer base using its platform directly. | 中 | SU001, SU002, SU005 |
| CU013 | Tempus’ oncology business publicly markets testing, trial matching, and care-pathway tools to thousands of oncologists and more than half of U.S. academic medical centers. | 中 | SU012 |
| CU014 | Tempus says 6.5K+ oncologists rely on its platform, with 45M+ records and 4.5K+ connected healthcare institutions. | 中 | SU012 |
| CU015 | Tempus reported $87.0 million in Q1 2026 data-and-applications revenue, indicating a mature commercial data platform that is far more scaled than anything Pathos publicly discloses. | 中 | SU011 |
| CU016 | The Tempus team page says the company has partnered with hundreds of biopharmaceutical companies, underscoring how much more mature its customer footprint is than Pathos' public footprint. | 中 | SU013 |
| CU017 | Flatiron says it works with hundreds of cancer centers and more than 20 top global developers of oncology therapeutics. | 中 | SU014 |
| CU018 | Foundation Medicine says it has delivered more than 1.5 million patient genomic profiling reports and supports more than half of approved U.S. CDx indications for NGS testing. | 中 | SU015 |
| CU019 | ConcertAI describes AI and real-world-data products for life-sciences customers and explicitly references life-science and healthcare clients. | 中 | SU016 |
| CU020 | PathAI says AISight is used by leading laboratories and research centers, indicating another oncology-adjacent competitor with clearer productized adoption language than Pathos uses publicly. | 中 | SU017 |
| CU021 | Owkin presents a patient-data network and AI agents for drug discovery and development, showing that Pathos is not alone in selling an AI-plus-patient-data future to pharma buyers. | 中 | SU018 |
| CU022 | Recursion publicly combines platform partnerships with an owned pipeline and claims more than 50 petabytes of proprietary data, providing another benchmark for Pathos customer expectations. | 高 | SU019, SU020 |
| CU023 | Caris publicly reports more than 1.0 million cases and 6.5 million tests, while its Q1 2026 revenue reached $216.2 million. | 高 | SU021, SU022 |
| CU024 | Compared with Tempus, Caris, Foundation, Flatiron, and ConcertAI, Pathos has far less public evidence of broad external customer penetration. | 中 | SU001, SU012, SU014, SU015, SU016, SU021, SU022 |
| CU025 | Pathos’ named external proof today is concentrated in Tempus and AstraZeneca rather than distributed across many buyers or users. | 中 | SU004, SU008, SU009 |
| CU026 | The strongest public proof of Pathos commercial utility currently shows Pathos as a customer of Tempus trial-enablement and data infrastructure, not as the vendor serving Tempus. | 中 | SU008, SU009, SU011 |
| CU027 | Novo Nordisk, Prelude, Rain, and Deuter are disclosed as transaction counterparties or acquisition targets, not as recurring customers of the PathOS platform. | 中 | SU005, SU006, SU007 |
| CU028 | Public materials do not disclose contract duration, renewal rates, NRR, GRR, churn, or customer satisfaction for any external Pathos relationship. | 中 | SU001, SU002, SU004, SU006, SU007 |
| CU029 | The lack of retention metrics makes it impossible to determine whether Pathos relationships are one-off transactions, durable programs, or expanding accounts. | 中 | SU004, SU007, SU008, SU009 |
| CU030 | Pathos does not publish pricing, package tiers, or procurement pathways for the platform on its public site. | 中 | SU001, SU002, SU003 |
| CU031 | The most plausible near-term expansion path is deeper pharma collaboration and more Pathos-sponsored trials rather than a rapid jump to broad clinical-provider adoption. | 中 | SU004, SU005, SU006, SU007, SU008 |
| CU032 | NCI says biomarker testing is important to precision medicine but not routine for most patients, which can slow the scaling of biomarker-led commercial workflows. | 中 | SU023 |
| CU033 | HHS guidance says even de-identified data retains some re-identification risk, raising diligence questions for any customer asked to rely on large patient-data networks. | 中 | SU024 |
| CU034 | The Springer review flags bias, privacy, and consent risks that can reduce buyer willingness to rely on AI-driven oncology decision systems without strong governance. | 中 | SU025 |
| CU035 | Fierce Biotech reported that Pathos did not disclose the participants in its Series D round, limiting one external signal of strategic customer or investor validation. | 中 | SU026 |
| CU036 | Tempus life sciences says it serves 5K+ connected institutions and 5M+ cancer patients in the TIME network, showing what a scaled oncology data and trial network looks like in this market. | 中 | SU010 |
| CU037 | Tempus life-sciences testimonials from Merck, Boehringer Ingelheim, and Verastem show the kind of named customer proof mature oncology data platforms can surface, which Pathos itself does not yet publish. | 中 | SU010 |
| CU038 | Pathos has enough public proof to say there is real external willingness to work with the company, but not enough to size the breadth, durability, or monetization of that demand. | 中 | SU004, SU008, SU009, SU026 |
| CU039 | Because Pathos is a private company with sparse commercial disclosures, customer concentration risk cannot yet be quantified by revenue share or account count from public evidence. | 低 | |
| CU040 | The current public customer story is therefore best described as enterprise-relationship proof without portfolio-level commercial transparency. | 中 | SU001, SU004, SU008, SU011, SU026 |
| CR001 | Pathos’ risk profile is shaped by its dependence on large external oncology data sets and partner infrastructure rather than solely by internal execution. | 中 | SR003, SR004, SR005, SR006, SR009 |
| CR002 | The PathOS platform claim depends on access to more than 200 petabytes of data and therefore raises data-rights, privacy, and governance risk if that access changes. | 高 | SR003, SR017 |
| CR003 | HHS states that even properly de-identified health information retains some residual risk of re-identification. | 中 | SR017 |
| CR004 | The Pathos-Tempus-AstraZeneca collaboration routes de-identified oncology data from Tempus into the foundation-model build. | 高 | SR004, SR005 |
| CR005 | Both the Pathos and Tempus announcements say the agreements include $200 million in fees to Tempus, creating supplier concentration and execution dependency. | 高 | SR004, SR005 |
| CR006 | The ASCO Post says legal standards for AI-related diagnostic and treatment errors in oncology remain unsettled. | 中 | SR018 |
| CR007 | The Springer review identifies bias, privacy, autonomy, and informed-consent issues as core risks for AI-driven oncology decision systems. | 中 | SR019 |
| CR008 | NCI says biomarker testing is important to precision medicine but is not routine care for most patients, limiting how quickly biomarker-led workflows can scale. | 中 | SR020 |
| CR009 | FDA says direct-to-consumer tests have varying levels of evidence supporting their claims and that some are not reviewed by FDA before being offered. | 中 | SR023 |
| CR010 | FDA says consumers should not make health-related decisions from DTC test results without first discussing them with a provider. | 中 | SR023 |
| CR011 | FDA maintains a standing recalls and safety-alert process because not all product risk is visible before launch, underscoring the need for post-market monitoring in regulated health products. | 中 | SR024 |
| CR012 | The Tempus TIME case study says early-phase oncology studies face delays from site activation and patient identification bottlenecks. | 中 | SR009 |
| CR013 | The same TIME case study shows that six of the first ten matched patients in the Pathos study were identified through Tempus, indicating external enrollment dependence. | 高 | SR009, SR011 |
| CR014 | Pathos said the pocenbrodib study is expected to enroll approximately 203 patients with mCRPC, a scale that still leaves material recruitment and efficacy threshold risk. | 高 | SR010, SR011 |
| CR015 | Urology Times reports the trial proceeds to later expansion only if acceptable safety and a minimum efficacy threshold are achieved, making early clinical underperformance a thesis-break risk. | 中 | SR011 |
| CR016 | The public Pathos pipeline still names only pocenbrodib and P-500, which leaves limited visible diversification if either program underperforms. | 中 | SR003 |
| CR017 | BioSpace reported grade 3 or higher thrombocytopenia, anemia, and fatigue in the earlier P-500 safety population. | 中 | SR013 |
| CR018 | The same report framed P-500 evidence as early phase 1 signal in specific subgroups, which keeps efficacy and reproducibility risk high. | 中 | SR013 |
| CR019 | Pathos said DO-2 came from a 28-patient phase 1 study, which is promising but still early enough to carry substantial replication and external-validation risk. | 高 | SR014, SR015 |
| CR020 | Pathos claims DO-2 had 100% tumor shrinkage in all evaluable MET exon 14 skipping NSCLC patients in that phase 1 study. | 高 | SR014, SR015 |
| CR021 | Pathos also claims a 5% peripheral edema rate for DO-2 versus 62% to 82% for competitors, but that comparison still requires broader confirmatory evidence. | 高 | SR014, SR015 |
| CR022 | Rain became a wholly owned subsidiary of Pathos in January 2024, creating integration and portfolio-prioritization risk across legacy assets. | 中 | SR016 |
| CR023 | Fierce Biotech noted that Pathos has not highlighted milademetan in more recent releases, suggesting reprioritization risk for acquired assets. | 中 | SR032 |
| CR024 | The SEC submissions feed shows Pathos has relied on repeated Form D fundraising across 2023, 2024, and 2025. | 高 | SR025, SR026, SR027, SR028 |
| CR025 | The 2025 Form D lists a total offering amount of about $400.0 million, with about $283.0 million sold and about $117.0 million remaining to be sold at filing time. | 中 | SR026 |
| CR026 | The 2024 Form D lists a total offering amount of about $62.0 million, consistent with the Series C announcement. | 高 | SR012, SR027 |
| CR027 | The 2023 Form D listed a roughly $40.0 million offering, showing that Pathos has required repeated external capital since launch. | 中 | SR028 |
| CR028 | Caris’ public scale — more than 1.0 million cases and $216.2 million of Q1 2026 revenue — shows that Pathos competes in a market where better-instrumented peers already exist. | 高 | SR029, SR030 |
| CR029 | Caris also highlights risks related to data security, patient privacy, and healthcare data-protection compliance, which are likely relevant to Pathos even if Pathos discloses less. | 中 | SR030 |
| CR030 | Caris’ S-1 says it will use emerging-growth-company exemptions from auditor attestation on internal control, underscoring governance risk that can persist even in scaled precision-oncology peers. | 中 | SR031 |
| CR031 | Tempus’ Q1 2026 release warns about evolving AI regulation, competition, IP protection, debt, and acquisition execution, all of which are category risks for Pathos as well. | 中 | SR007 |
| CR032 | The Tempus leadership page surfaces dedicated privacy, security, and legal executives, while Pathos’ public site surfaces far less governance detail. | 中 | SR002, SR008 |
| CR033 | Pathos’ public site does not expose a trust center, SOC report, or model-governance package. | 中 | SR001, SR002, SR003 |
| CR034 | NCI estimates 2,041,910 new U.S. cancer cases in 2025 and national cancer-care expenditures of $208.9 billion in 2020, which makes product, data, and safety mistakes materially consequential. | 中 | SR021 |
| CR035 | SEER estimates 2,114,850 new cancer cases and 626,140 deaths in 2026, reinforcing the high clinical stakes of any model-guided oncology workflow. | 中 | SR022 |
| CR036 | Tempus life sciences markets 8.5 million research records and more than 5,000 connected institutions, which means Pathos’ strategic model build sits on top of a partner with much larger operational scale than Pathos itself. | 中 | SR006 |
| CR037 | The strongest Pathos operational proof currently runs through Tempus, making the company vulnerable to partner friction in data access, enrollment support, or commercial terms. | 中 | SR004, SR005, SR009, SR013 |
| CR038 | Because Pathos has not publicly disclosed platform uptime, validation thresholds, or incident history, operational and quality risk cannot be reduced below medium from public evidence. | 中 | SR001, SR002, SR003, SR033 |
| CR039 | Open questions remain around the exact contractual data-rights chain and contingency plan if the Tempus relationship changes. | 低 | |
| CR040 | Open questions also remain around which internal governance, safety, and model-review controls Pathos requires before platform output informs portfolio decisions or trial design. | 低 | |
| CV001 | Using the disclosed Series D post-money mark of about $1.6 billion, Pathos was valued at roughly the same level as Recursion Pharmaceuticals' late-May 2026 public market capitalization. | 中 | SV001, SV013 |
| CV002 | Pathos announced a $62 million Series C financing in October 2024 at a $600 million post-money valuation. | 高 | SV002, SV010 |
| CV003 | The disclosed post-money valuation therefore increased by about 2.7x from the Series C to the Series D in roughly seven months. | 高 | SV001, SV002 |
| CV004 | Independent coverage said the identities of the Series D participants were not disclosed publicly. | 中 | SV003 |
| CV005 | The 2025 Form D lists a total offering amount of about $400.0 million, with about $283.0 million sold and about $117.0 million remaining at filing time. | 中 | SV009 |
| CV006 | The 2024 Form D lists a total offering amount of about $62.0 million, aligning with the Series C announcement. | 高 | SV002, SV010 |
| CV007 | The 2023 Form D lists a roughly $40.0 million offering, showing Pathos began building its capital base well before the 2024 and 2025 rounds. | 中 | SV011 |
| CV008 | Public Pathos financing announcements and SEC filings together support at least about $467 million of announced or offered capital across 2023 through 2025. | 中 | SV001, SV002, SV009, SV010, SV011 |
| CV009 | Pathos does not publicly disclose revenue, ARR, gross margin, or cash runway in the reviewed public materials. | 中 | SV001, SV002, SV025 |
| CV010 | Because Pathos has no public revenue base, standard revenue-multiple or EBITDA-multiple valuation methods cannot be anchored to reported financials. | 中 | SV001, SV002, SV025 |
| CV011 | Tempus AI had a market cap of about $8.29 billion in late May 2026. | 中 | SV012 |
| CV012 | Tempus reported Q1 2026 revenue of $348.1 million and 2026 revenue guidance of $1.59 billion to $1.60 billion. | 中 | SV022 |
| CV013 | Recursion Pharmaceuticals had a market cap of about $1.59 billion in late May 2026. | 中 | SV013, SV017 |
| CV014 | Recursion’s enterprise value on StockAnalysis was about $1.02 billion. | 中 | SV017 |
| CV015 | Schrödinger had a market cap of about $0.99 billion in late May 2026. | 中 | SV014, SV018 |
| CV016 | Relay Therapeutics had a market cap of about $2.90 billion in late May 2026. | 中 | SV015 |
| CV017 | Absci had a market cap of about $795 million in late May 2026. | 中 | SV016 |
| CV018 | Caris reported Q1 2026 revenue of $216.2 million and reaffirmed full-year 2026 revenue guidance of $1.0 billion to $1.02 billion. | 中 | SV019 |
| CV019 | Caris’ 2025 10-K said the aggregate market value of non-affiliate equity was approximately $3.87 billion as of June 30, 2025. | 中 | SV021 |
| CV020 | Relative to public comps, Pathos’ $1.6 billion private valuation is below Tempus and Caris but roughly in line with Recursion and above Schrödinger and Absci. | 中 | SV001, SV012, SV013, SV014, SV016, SV021 |
| CV021 | That means the Series D priced Pathos like a public AI-biotech platform before the company has disclosed any public revenue base. | 中 | SV001, SV002, SV012, SV013, SV014, SV016 |
| CV022 | The Tempus and AstraZeneca collaboration adds strategic validation but also commits Pathos to a large external fee stream to Tempus. | 高 | SV006, SV007 |
| CV023 | Independent coverage says Pathos is building an oncology foundation model spanning clinical, molecular, and imaging data. | 中 | SV004 |
| CV024 | The public bull case rests on turning that model plus the asset portfolio into repeatable clinical and portfolio wins, not on current revenue. | 中 | SV001, SV002, SV004, SV025 |
| CV025 | The public bear case rests on early-clinical failure, partner concentration, valuation compression, and limited transparency on underlying economics. | 中 | SV003, SV006, SV009, SV025 |
| CV026 | Fierce reported that P-500 remained a planned future trial and that Rain’s milademetan had not been highlighted in more recent releases. | 中 | SV003, SV005 |
| CV027 | GenomeWeb reported that Pathos was founded by Tempus executives and had recently signed the Tempus/AstraZeneca collaboration before the Series D. | 中 | SV004 |
| CV028 | Pharmaphorum reported that P-500 had completed phase 1 testing with a mid-stage trial planned, which supports upside but keeps program risk squarely pre-commercial. | 中 | SV005, SV027 |
| CV029 | Caris and Tempus show that revenue-scale peers in this category already publish volumes, revenue, and guidance that Pathos does not. | 中 | SV019, SV022, SV025 |
| CV030 | Public-market performance among AI-biotech peers is volatile: Tempus, Recursion, and Schrödinger all showed year-over-year market-cap declines in the fetched market data. | 中 | SV012, SV013, SV014 |
| CV031 | Absci and Relay show the opposite direction, proving that public comp dispersion is wide and that the right valuation framework for Pathos must be scenario-based. | 中 | SV015, SV016 |
| CV032 | A scenario-based range is more defensible than a single point estimate because Pathos has little public financial telemetry and large outcome dispersion. | 中 | SV001, SV002, SV012, SV013, SV014, SV015, SV016 |
| CV033 | A reasonable public bear range is roughly $0.7 billion to $1.0 billion, anchored on smaller public AI-biotech platforms with lower or no near-term commercial proof. | 中 | SV014, SV016, SV018 |
| CV034 | A reasonable public base range is roughly $1.4 billion to $1.8 billion, centered around the current private mark and the public Recursion comparison. | 中 | SV001, SV013, SV017 |
| CV035 | A reasonable public bull range is roughly $2.5 billion to $3.5 billion if Pathos converts the platform narrative into credible multi-asset clinical progress and more partner wins. | 中 | SV001, SV004, SV006, SV007 |
| CV036 | Given the lack of public revenue and cap-table detail, the prudent investment stance is track rather than buy. | 中 | SV001, SV002, SV003, SV025 |
| CV037 | Confidence in that stance is only medium because the public record is strong on financing and strategy but weak on economics and validated outcomes. | 中 | SV001, SV002, SV004, SV025 |
| CV038 | The valuation stance is stretched rather than attractive because the private mark already embeds significant execution success relative to current public proof. | 中 | SV001, SV013, SV014, SV016, SV021 |
| CV039 | Key missing diligence items include the exact cash balance, cap table, investor rights, preference stack, budget by asset, and any secondary component of the Series D. | 低 | |
| CV040 | The most important thesis-break triggers are a failure to generate credible pocenbrodib proof, disruption in the Tempus relationship, or a down-round driven by fading platform confidence. | 中 | SV003, SV006, SV009, SV030 |
| CV041 | Until Pathos discloses more financial and operating data, the public evidence supports continued monitoring and tighter entry discipline rather than aggressive underwriting. | 中 | SV001, SV002, SV003, SV025 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | Pathos AI | Pathos AI Homepage | Pathos is an AI-enabled, clinical-stage biotech building a new development engine for precision oncology through multimodal data, biological modeling, and biopharma partnerships. |
| SO002 | Pathos AI | Pathos AI About Us | Pathos was founded in 2022 by Eric Lefkofsky and Ryan Fukushima, after they realized that AI could have a profound impact on drug discovery and development. |
| SO003 | Pathos AI | Pathos AI Platform — PathOS Architecture | |
| SO004 | Pathos AI | Pathos AI Pipeline | |
| SO005 | Pathos AI | Pathos AI Secures $365 Million in Series D Financing to Advance Oncology Drug Development Through AI | Pathos AI, (www.pathos.com), a leading AI-driven biotech company applying cutting-edge artificial intelligence to drug development, today announced its $365 million Series D financing, bringing its post-money valuation to approximately $1.6 billion. |
| SO006 | Pathos AI | Pathos AI Closes $62M Oversubscribed Series C Round of Financing | The Series C financing round was led by New Enterprise Associates (NEA) with participation from Revolution Growth and other existing insiders. This latest round … was completed at a $600 million post money valuation, bringing the three-year-old company's total funding to $102M. |
| SO007 | Pathos AI | Pathos Launches Precision Oncology Pipeline with License of First Phase I Program — CBP/P300 Inhibitor | Pathos AI, Inc. … announced today that it has entered into a worldwide license agreement to develop FT-7051, a small molecule CBP/p300 inhibitor program from Novo Nordisk as Pathos' first clinical-stage asset in its pipeline. |
| SO008 | Pathos AI | Pathos AI Completes Acquisition of Rain Oncology | Pathos AI, Inc. … today announced that it has … successfully completed its tender offer to acquire all outstanding shares of the common stock of Rain Oncology Inc. … for $1.16 per share in cash plus one contingent value right per share. |
| SO009 | Pathos AI | Pathos AI Doses First Patient in Phase 1b/2a Clinical Trial of Pocenbrodib | Pathos AI … announced the first patient has been dosed in the Company's Phase 1b/2a clinical trial evaluating pocenbrodib, a CBP/p300 inhibitor … in patients with metastatic castration-resistant prostate cancer (mCRPC), (P300-02-001, NCT06785636). |
| SO010 | Pathos AI | Pathos Signs Strategic Agreements with AstraZeneca and Tempus to Develop the Largest Multimodal Foundation Model in Oncology | The agreements include $200 million in data licensing and model development fees to Tempus. |
| SO011 | Pathos AI | Pathos AI Acquires Majority Stake in DeuterOncology | DO-2 is one of four major portfolio decisions made through Foundry in Q1 2026 alone. |
| SO012 | U.S. Securities and Exchange Commission | SEC EDGAR Submissions — Pathos AI, Inc. (CIK 0001967854) | |
| SO013 | U.S. Securities and Exchange Commission | Pathos AI Form D — Series D (2025-05-01) | |
| SO014 | U.S. Securities and Exchange Commission | Pathos AI Form D — Series C (2024-11-06) | |
| SO015 | U.S. Securities and Exchange Commission | Pathos AI Form D — First Filing (2023-03-02) | |
| SO016 | Pathos AI | Pathos AI Careers Page | |
| SO017 | Fierce Biotech | Pathos AI Secures $365M Series D to Fund Trial of Novo Nordisk Solid Tumor Drug | Pathos AI has secured $365 million in a series D raise as the artificial-intelligence-powered biotech looks to fund trials of solid tumor drugs sourced from Novo Nordisk and Prelude Therapeutics. |
| SO018 | GenomeWeb | Pathos AI Raises $365M Series D Financing to Advance AI Oncology Drug Development | This article has been updated to correct that Pathos AI is not a spinoff of Tempus AI but was founded by Tempus AI executives. |
| SO019 | Pharmaphorum | Pathos AI's Big $365M Series D and Other Bio Financings | |
| SO020 | Pharmaceutical Technology | Pathos AI Raises $365M in Series D Funding | |
| SO021 | BioSpace | Pathos AI Acquires Majority Stake in DeuterOncology | In a Phase 1 study of 28 patients, DO-2 demonstrated 100% tumor shrinkage in all evaluable MET exon 14 skipping NSCLC patients (10/10). It also demonstrated a superior safety profile with zero Grade 4 adverse events, and a peripheral edema rate of just 5%. |
| SO022 | BioSpace | Pathos Expands Pipeline with Worldwide License of Phase 2-Ready Brain-Penetrant PRMT5 Inhibitor | Out of 16 patients with high-grade glioma with isocitrate dehydrogenase mutations (IDH+) in the Phase 1 trial, two confirmed complete responses (CR) were observed. At last follow-up, 1 response is ongoing and has lasted 31.0 months. |
| SO023 | BioSpace | Pathos AI Doses First Patient in Phase 1b/2a Clinical Trial of Pocenbrodib | |
| SO024 | BioPharma Trend | Pathos AI Acquires Brain-Penetrant PRMT5 Inhibitor for Precision Cancer Therapy | |
| SO025 | Tempus AI | Tempus Signs Expanded Strategic Agreements with AstraZeneca and Pathos to Develop the Largest Multimodal Foundation Model in Oncology | The agreements include $200 million in data licensing and model development fees to Tempus. |
| SO026 | Urology Times | Trial Launches of CBP/P300 Inhibitor in mCRPC | The first patient has been dosed in a phase 1b/2a trial (NCT06785636) of pocenbrodib, a CBP/p300 inhibitor, as a monotherapy and in combination with abiraterone acetate (Zytiga), olaparib (Lynparza), or 177Lu-PSMA-617 (Pluvicto) in patients with metastatic castration-resistant prostate cancer (mCRPC). |
| SO027 | ASCO Post | Understanding the Legal and Ethical Challenges AI Poses in Oncology | Although artificial intelligence tools offer promising advancements in diagnosis and treatment, their legal implications are evolving and uncertain. |
| SO028 | Springer / Current Oncology Reports | Artificial Intelligence in Oncology: Current Status and Future Directions | |
| SO029 | Craft.co | Pathos AI Locations | |
| SO030 | CompaniesBio | Pathos AI, Inc. — SEC Filings Summary | |
| SO031 | Fierce Biotech | Tempus AI in Line for $200M as AstraZeneca and Pathos Deal to Develop Cancer Model | |
| SO032 | Chicago Business | Pathos AI Raises $365 Million, Touts $1.6 Billion Valuation | |
| SO033 | BioPharma Trend | Tempus, AstraZeneca and Pathos Partner to Build Oncology Foundation Model Using Multimodal Data | |
| SO034 | ClinicalTrials.gov | NCT06785675 — Pathos AI Clinical Trial Registry | |
| SM001 | IQVIA Institute | Global Oncology Trends 2025 | Cancer medicine spending at list prices rose to $252Bn globally in 2024 and is expected to reach $441Bn by 2029. |
| SM002 | MarketsandMarkets | AI in Oncology Market, Drug Discovery Technologies Market, and Precision Diagnostics Market Reports | |
| SM003 | Mordor Intelligence | Real-World Evidence Solutions Market Analysis | By therapeutic area, oncology commanded 34.65% of the real-world evidence solutions market share in 2025. |
| SM004 | American Cancer Society | Cancer Facts & Figures 2025 | For 2025, there will be an estimated 2,041,910 new cancer cases diagnosed and 618,120 cancer deaths in the United States. |
| SM005 | World Health Organization | Cancer Fact Sheet | Cancer is a leading cause of death worldwide, accounting for nearly 10 million deaths in 2022. |
| SM006 | Centers for Disease Control and Prevention | Highlights from 2025 U.S. Cancer Statistics | U.S. Cancer Statistics has achieved 100% coverage of new cancers and deaths reported in all 50 states and the District of Columbia over a 20-year period (2003 to 2022). During this period, 36.7 million new cancer cases were reported. |
| SM007 | Morningstar / Business Wire | Tempus Reports First Quarter 2026 Results | Revenue of $348.1 million, up 36.1% year-over-year. |
| SM008 | Caris Life Sciences, Inc. | Caris Life Sciences Annual Report (10-K) for Fiscal Year 2025 | |
| SM009 | Morningstar / PR Newswire | Caris Life Sciences Reports First Quarter 2026 Financial Results | Reported total revenue of $216.2 million, an increase of 79% over the corresponding prior year period. |
| SM010 | Business Wire | Tempus Enters Multi-Year Strategic Collaboration With Boehringer Ingelheim to Advance Its Cancer Pipeline | Tempus announced a multi-year strategic collaboration with Boehringer Ingelheim to advance its cancer pipeline. |
| SM011 | ConcertAI | ConcertAI — AI-Powered Oncology Real-World Data | |
| SM012 | MedCity News | ConcertAI Lands $150M Round Led by Goldman Sachs Asset Management at $1.9B Valuation | |
| SM013 | Springer / Current Oncology Reports | Artificial Intelligence in Oncology: Legal and Ethical Implications | The legal landscape surrounding AI-driven decision-making in oncology remains unclear, posing significant medico-legal risks. |
| SM014 | Pathos AI, Inc. | Pathos AI Secures $365 Million in Series D Financing | |
| SM015 | Allied Market Research | Next-Generation Sequencing Market Report | |
| SM016 | Grand View Research | AI in Clinical Trials Market Report | |
| SM017 | Flatiron Health | About Flatiron Health | |
| SM018 | Pathos AI, Inc. | Pathos AI Signs Strategic Agreements with AstraZeneca and Tempus | |
| SM019 | Pathos AI, Inc. | PathOS Platform | |
| SM020 | Tempus AI, Inc. | Tempus Signs Expanded Strategic Agreements with AstraZeneca and Pathos | |
| SM021 | FierceBiotech | Tempus AI in Line for $200M from AstraZeneca, Pathos Deal to Develop Cancer Model | |
| SM022 | FierceBiotech | Pathos AI Secures $365M Series D to Fund Trial of Novo Nordisk Solid Tumor Drug | |
| SM023 | HIT Consultant | Syapse Lands $68M to Expand Global Precision Oncology Data Sharing Network | Syapse announced $68 million to expand its global precision oncology data sharing network. |
| SM024 | Roche Diagnostics | Comprehensive Genomic Profiling (CGP) Solutions | Comprehensive genomic profiling helps identify clinically relevant cancer biomarkers and support precision treatment decisions. |
| SM025 | U.S. Food and Drug Administration | Real-World Evidence | FDA has a long history of using what we currently call real-world data (RWD) and real-world evidence (RWE) to monitor and evaluate the postmarket safety of approved drugs. |
| SM026 | Reuters | Caris Life Sciences Discloses Rise in Revenue in US IPO Filing | |
| SM027 | Owkin | What is Federated Learning? | Federated learning enables AI training across institutions without centralizing raw data. |
| SM028 | Vivli | Tempus AI - Vivli | |
| SM029 | Caris Life Sciences, Inc. | About | Caris Life Sciences | |
| SM030 | GenomeWeb | Pathos AI Raises $365M Series D Financing to Advance AI Oncology Drug Development | |
| SP001 | Pathos | About Us | Pathos | |
| SP002 | Pathos | Pipeline | Pathos | |
| SP003 | GlobeNewswire | Recursion and Exscientia Enter Definitive Agreement to Create a Global Technology-Enabled Drug Discovery Leader with End-to-End Capabilities | |
| SP004 | Nasdaq | Recursion and Exscientia, two leaders in the AI drug discovery space, have officially combined | |
| SP005 | Insilico Medicine | Pipeline | Insilico Medicine | |
| SP006 | PR Newswire | Insilico Medicine Announces Global R&D Collaboration with Lilly | |
| SP007 | BenevolentAI | BenevolentAI provides an update on its business priorities | |
| SP008 | BenevolentAI | BenevolentAI Signs Strategic Collaboration with Merck | |
| SP009 | BenevolentAI | BenevolentAI unveils major strategic overhaul with return to original mission | |
| SP010 | BenevolentAI | Proposed Delisting via Merger of BenevolentAI into Osaka Holdings S.à r.l. and Publication of Notice of Extraordinary General Meeting | |
| SP011 | Schrödinger | Life science - Schrödinger | |
| SP012 | Schrödinger | Pipeline of Collaborative & Proprietary Drug Discovery Programs - Schrödinger | |
| SP013 | Schrödinger | Schrödinger Announces Multi-Target Collaboration and Expanded Software Licensing Agreement with Novartis | |
| SP014 | Relay Therapeutics | Our Science - Relay Therapeutics | |
| SP015 | Relay Therapeutics | Dynamo Platform - Relay Therapeutics | |
| SP016 | Elevar Therapeutics | Elevar Therapeutics and Relay Therapeutics Announce Exclusive Global Licensing Agreement for Lirafugratinib | |
| SP017 | Valo Health | This is Intelligent Health | |
| SP018 | Business Wire | Valo Health Appoints Karin Conde-Knape, Ph.D., as Chief Scientific Officer | |
| SP019 | Nasdaq | Ikena Oncology and Inmagene Biopharmaceuticals Announce Agreement for Merger and Private Placement | |
| SP020 | Nasdaq | Inmagene Biopharmaceuticals Announces Completion of Merger with Ikena Oncology and Concurrent Private Placement of $75 Million | |
| SP021 | Boundless Bio | Boundless Bio | |
| SP022 | Absci | Our Pipeline | Absci | |
| SP023 | Business Wire | Tempus Molecular Profiling Integration Now Available Through Flatiron’s OncoEMR® | |
| SP024 | TechCrunch | Billionaire Groupon founder Eric Lefkofsky is back with another IPO: AI health tech Tempus | |
| SP025 | PR Newswire | Caris Life Sciences and Flatiron Health Partner to Create Transformational Real-World Data Offering | |
| SP026 | PR Newswire | ConcertAI and Caris Life Sciences Announce Strategic Agreement with AbbVie to Accelerate Oncology Pipeline and Clinical Trials | |
| SP027 | Business Wire | ConcertAI Launches New Generative and Agentic AI-Powered Precision Suite™ Accelerating Oncology Insights and Actions for Healthcare and Life Sciences | |
| SP028 | Business Wire | Flatiron Health Brings 25+ Research Acceptances and Next-Generation Capabilities to ASCO 2026 | |
| SP029 | Roche | Roche | Foundation medicine | |
| SP030 | Business Wire | Owkin Announces Partnership with AstraZeneca to Develop an AI gBRCA Pre-Screen Solution for Breast Cancer | |
| SP031 | PR Newswire | OWKIN Integrates 10x Genomics Spatial Omics and Single-Cell Technologies to the MOSAIC Study | |
| SP032 | Sanofi | Partner Spotlight: Owkin and Sanofi embrace the “golden era” of precision medicine with AI | |
| SI001 | Pathos AI | Pathos AI Secures $365 Million in Series D Financing to Advance Oncology Drug Development Through AI | Pathos AI today announced its $365 million Series D financing, bringing its post-money valuation to approximately $1.6 billion. |
| SI002 | Pathos AI | Pathos AI Closes $62M Oversubscribed Series C Round of Financing to Accelerate its Platform Approach to Drug Development | The Series C financing round was led by New Enterprise Associates (NEA) and was completed at a $600 million post money valuation. |
| SI003 | Pathos AI | Pathos Launches Precision Oncology Pipeline With License of First Phase I Program, a CBP/p300 Inhibitor | Pathos announced a worldwide license agreement to develop FT-7051 as its first clinical-stage asset. |
| SI004 | Pathos AI | Pathos Signs Strategic Agreements with AstraZeneca and Tempus to Develop the Largest Multimodal Foundation Model in Oncology | The agreements include $200 million in data licensing and model development fees to Tempus. |
| SI005 | U.S. Securities and Exchange Commission | SEC EDGAR submissions for Pathos AI, Inc. (CIK 0001967854) | |
| SI006 | U.S. Securities and Exchange Commission | Pathos AI Form D filing (2025-05-01) | |
| SI007 | U.S. Securities and Exchange Commission | Pathos AI Form D filing (2024-11-06) | |
| SI008 | U.S. Securities and Exchange Commission | Pathos AI Form D filing (2023-03-02) | |
| SI009 | Business Wire | Pathos AI Completes Acquisition of Rain Oncology | Rain stockholders were to receive $1.16 per share in cash plus contingent value rights. |
| SI010 | Business Wire | Variational AI Announces Oversubscribed $5.5 Million Financing to Launch Foundation Model for Small Molecule Drug Discovery | Variational AI completed an oversubscribed $US5.5 million seed extension round. |
| SI011 | PR Newswire | Insilico Medicine Secures $110 Million Series E Financing to Advance AI-Driven Drug Discovery Innovation | Insilico Medicine announced that it had secured $110 million of Series E financing. |
| SI012 | PR Newswire | Isomorphic Labs secures $2.1 Billion funding to scale its AI drug design engine | Isomorphic Labs announced it had raised $2.1 billion in Series B funding. |
| SI013 | McKinsey & Company | Small but mighty: Priming biotech first-time launchers to compete with established players | Roughly 40 percent of new assets submitted for FDA approval between 2018 and 2023 came from companies with little to no commercialization experience. |
| SI014 | Nature Biotechnology | Biotech financing: darkest before the dawn | After a difficult three years, biotech financing may slowly be returning to health. |
| SI015 | Nature | Biotech trends driving the deals of 2025 | The top licensing partnerships of 2025 highlight continued reliance on AI for drug discovery. |
| SI016 | Nature | This AI method could turbocharge the hunt for new medicines | An AI model trained on complex data from human cells could provide a shortcut in the race to develop new drugs. |
| SI017 | Schrödinger | Schrödinger Reports First Quarter 2026 Financial Results | Cash, cash equivalents, restricted cash and marketable securities were $406 million at the end of the first quarter of 2026. |
| SI018 | Schrödinger | Computational Platform for Molecular Discovery & Design | Schrödinger describes an industry-leading computational platform used across biotech and pharmaceutical discovery. |
| SI019 | Relay Therapeutics | Relay Therapeutics Reports First Quarter 2026 Financial Results and Corporate Updates | As of March 31, 2026, cash, cash equivalents and investments totaled $642.1 million. |
| SI020 | Relay Therapeutics | Relay Therapeutics homepage | Relay says it is advancing a pipeline of therapeutic candidates with an initial focus on precision oncology and genetic disease. |
| SI021 | Relay Therapeutics | Relay Therapeutics Reports First Quarter 2025 Financial Results and Corporate Updates | Relay reported approximately $710 million in cash, cash equivalents and investments at the end of Q1 2025 and said runway was extended into 2029. |
| SI022 | Insilico Medicine | Insilico Medicine Announces 2025 Annual Results, Redefining Value Delivery in AI-Powered Drug Discovery | The company's cash and bank balances were US$393.3 million as of December 31, 2025 and 2025 revenue was US$56.24 million. |
| SI023 | Insilico Medicine | Investor Relations | Insilico Medicine | Insilico maintains a public investor-relations surface with governance, listing documents, presentations, and announcements. |
| SI024 | Insilico Medicine | Pipeline | Insilico Medicine | Insilico's pipeline page states 40-plus total programs and 13 pipelines that received IND approval. |
| SI025 | PubMed | Conducting clinical trials-costs, impacts, and the value of clinical trials networks: A scoping review | The review highlights that delayed activation, poor accrual, and organizational know-how all affect clinical-trial cost and value. |
| SI026 | PubMed | Cost-effectiveness as an outcome in randomized clinical trials | Economic outcomes can be measured alongside clinical outcomes in randomized trials, but doing so raises methodological and data-collection challenges. |
| SI027 | PubMed Central | What drives cancer clinical trial accrual? An empirical analysis of studies leading to FDA authorisation (2015-2020) | The study examines which design factors affect accrual rates in cancer trials that supported FDA approvals. |
| SI028 | JAMA Network Open | Costs of Drug Development and Research and Development Intensity in the US | The study estimated mean drug-development cost at $172.7 million before accounting for higher failed-program and capital-cost scenarios. |
| SI029 | Pathos AI | Pathos AI Acquires Majority Stake in DeuterOncology to Advance Next-Generation MET Inhibitor Identified by Pathos Foundry Platform | DO-2 is one of four major portfolio decisions made through Foundry in Q1 2026 alone. |
| SE001 | Pathos AI | PathOS Platform | We have access to >200 petabytes of multimodal oncology data linked to patient outcomes. |
| SE002 | Pathos AI | Pathos homepage | Pathos is an AI-enabled, clinical-stage biotech building a new development engine for precision oncology through multimodal data, biological modeling, and biopharma partnerships. |
| SE003 | Pathos AI | About Pathos | Today, Pathos uses this model to identify the right patients, design adaptive trials, and run them with small AI-enabled teams. |
| SE004 | Pathos AI | Pipeline | We’re advancing a pipeline built on precision, where each program teaches the next, and every success compounds value. |
| SE005 | Pathos AI | Careers | Legitimate emails from our recruitment team will only come from an official @pathos.com or @ats.rippling.com email address. |
| SE006 | Pathos AI | Pathos signs strategic agreements with AstraZeneca and Tempus | The agreements include $200 million in data licensing and model development fees to Tempus. |
| SE007 | Tempus AI | Tempus signs expanded strategic agreements with AstraZeneca and Pathos | Tempus’ de-identified oncology data will be used to build the foundation model. |
| SE008 | Tempus AI | Life sciences | 8.5M+ de-identified research records |
| SE009 | Pathos AI | Pathos launches precision oncology pipeline with license of first phase I program | Pathos obtains worldwide rights from Novo Nordisk for the development of CBP/p300 inhibitor, FT-7051. |
| SE010 | Pathos AI | Pathos AI doses first patient in phase 1b/2a clinical trial of pocenbrodib | The study is expected to enroll approximately 203 patients with mCRPC. |
| SE011 | Urology Times | Trial launches of CBP/p300 inhibitor in mCRPC | In total, the trial plans to enroll 203 adult patients with mCRPC across 3 clinical trial sites in the US. |
| SE012 | Pathos AI | Pathos closes $62M Series C financing | P-500, a phase-II ready, brain-penetrant PRMT5 inhibitor |
| SE013 | Pathos AI | Pathos secures $365 million Series D financing | Pathos is developing the largest multimodal foundation model in oncology. |
| SE014 | Fierce Biotech | Pathos AI secures $365M series D to fund trials | its other “key priority” this year is to launch a trial of P-500 |
| SE015 | GenomeWeb | Pathos AI raises $365M series D financing | Pathos claims to be developing the largest oncology foundation model to date, which combines clinical, molecular, and imaging data. |
| SE016 | Pharmaphorum | Pathos AI’s big $365m series D | P-500 ... has completed phase 1 testing with a mid-stage trial planned. |
| SE017 | Pathos AI | Pathos acquires majority stake in DeuterOncology | In a Phase 1 study of 28 patients, DO-2 demonstrated 100% tumor shrinkage in all evaluable MET exon 14 skipping NSCLC patients (10/10). |
| SE018 | BioSpace | Pathos AI acquires majority stake in DeuterOncology | peripheral edema rate of just 5% (versus 62-82% for competitors) |
| SE019 | Pathos AI | Pathos AI completes acquisition of Rain Oncology | Rain became a wholly owned subsidiary of Pathos. |
| SE020 | BioSpace | Pathos expands pipeline with worldwide license of PRMT5 inhibitor | Out of 16 patients with high-grade glioma with isocitrate dehydrogenase mutations (IDH+), two confirmed complete responses (CR) were observed. |
| SE021 | HHS Office for Civil Rights | Guidance regarding methods for de-identification of protected health information | Both methods, even when properly applied, yield de-identified data that retains some risk of identification. |
| SE022 | National Cancer Institute | Biomarker testing for cancer treatment | Biomarker testing is an important part of precision medicine. |
| SE023 | The ASCO Post | Understanding the legal and ethical challenges AI poses in oncology | Although artificial intelligence tools offer promising advancements in diagnosis and treatment, their legal implications are evolving and uncertain. |
| SE024 | Current Treatment Options in Oncology | Ethical, legal and informed consent challenges for AI-driven therapeutic decision-making in oncology | These models are only as effective as the data used to train them. |
| SE025 | Tempus AI | Accelerating a phase 1 oncology trial: The TIME Network’s impact on patient enrollment | Of the 10 patients enrolled on this trial to date, Tempus helped identify 6 matches through the TIME network. |
| SE026 | Tempus AI | Tempus reports first quarter 2026 results | Data and Applications revenue generated $87.0 million of revenue, representing 40.5% year-over-year growth, with Insights growing 44.1%. |
| SE027 | Tempus AI | Tempus Oncology | Trusted by thousands of oncologists. |
| SE028 | Recursion | Recursion home | Over the last decade, we have generated and aggregated one of the largest fit-for-purpose proprietary biological and chemical datasets in the world — >50 petabytes. |
| SE029 | Owkin | Owkin home | We use patient data to drive real-world impact, connecting research to care. |
| SU001 | Pathos AI | Pathos homepage | Pathos is an AI-enabled, clinical-stage biotech building a new development engine for precision oncology through multimodal data, biological modeling, and biopharma partnerships. |
| SU002 | Pathos AI | About Pathos | Partnering with pharma, biotech, academia, and investors to bring therapies to patients. |
| SU003 | Pathos AI | PathOS Platform | Sprint translates them into faster clinical learning. |
| SU004 | Pathos AI | Pathos signs strategic agreements with AstraZeneca and Tempus | The agreements include $200 million in data licensing and model development fees to Tempus. |
| SU005 | Pathos AI | Pathos AI doses first patient in phase 1b/2a clinical trial of pocenbrodib | The study is expected to enroll approximately 203 patients with mCRPC. |
| SU006 | Pathos AI | Pathos closes $62M Series C financing | By pairing the right patient selection strategies with great oncology therapeutics, Pathos believes it can partner with biopharma to usher in the next era of precision medicine. |
| SU007 | Pathos AI | Pathos secures $365 million Series D financing | The proceeds will support advancement of the company’s clinical-stage pipeline and continued investment in its proprietary AI Foundation Model purpose-built for oncology. |
| SU008 | Tempus AI | Accelerating a phase 1 oncology trial: The TIME Network’s impact on patient enrollment | Of the 10 patients enrolled on this trial to date, Tempus helped identify 6 matches through the TIME network. |
| SU009 | Tempus AI | Tempus signs expanded strategic agreements with AstraZeneca and Pathos | Tempus’ de-identified oncology data will be used to build the foundation model. |
| SU010 | Tempus AI | Life sciences | 5K+ connected healthcare institutions |
| SU011 | Tempus AI | Tempus reports first quarter 2026 results | Data and Applications revenue generated $87.0 million of revenue, representing 40.5% year-over-year growth, with Insights growing 44.1%. |
| SU012 | Tempus AI | Tempus Oncology | Trusted by thousands of oncologists |
| SU013 | Tempus AI | Our Team | Ryan has scaled Tempus from startup to public company... partnering with hundreds of biopharmaceutical companies. |
| SU014 | Flatiron Health | About Flatiron | We partner with hundreds of cancer centers, 20+ top global developers of oncology therapeutics. |
| SU015 | Foundation Medicine | Foundation Medicine home | Over 1.5 Million Patient Comprehensive Genomic Profiling reports delivered. |
| SU016 | ConcertAI | ConcertAI home | Our New Precision Suite of AI Products & Solutions |
| SU017 | PathAI | PathAI home | AISight is a cloud-native, open platform enterprise workflow solution used by the world's leading laboratories and research centers. |
| SU018 | Owkin | Owkin home | We start with patient data to drive real-world impact, connecting research to care. |
| SU019 | Recursion | Recursion home | Our strategic partnerships allow us to accelerate the discovery of new medicines and expand our potential impact on patients via AI-drug discovery. |
| SU020 | Recursion | Pipeline | REC-4881 is an orally bioavailable ... inhibitor being developed to reduce polyp burden and progression to adenocarcinoma. |
| SU021 | Caris Life Sciences | Caris home | 1.0+ Million Cases |
| SU022 | Caris Life Sciences | Caris Life Sciences reports first quarter 2026 financial results | Reported total revenue of $216.2 million |
| SU023 | National Cancer Institute | Biomarker testing for cancer treatment | Biomarker testing is an important part of precision medicine. |
| SU024 | HHS Office for Civil Rights | Guidance regarding methods for de-identification of protected health information | Both methods, even when properly applied, yield de-identified data that retains some risk of identification. |
| SU025 | Current Treatment Options in Oncology | Ethical, legal and informed consent challenges for AI-driven therapeutic decision-making in oncology | Bias in AI models remains a major ethical issue. |
| SU026 | Fierce Biotech | Pathos AI secures $365M series D to fund trials | Thursday's release didn't include the identities of the participants in the series D round. |
| SU027 | Urology Times | Trial launches of CBP/p300 inhibitor in mCRPC | In total, the trial plans to enroll 203 adult patients with mCRPC across 3 clinical trial sites in the US. |
| SU028 | Syapse | Syapse home | syapse.com - Website Moved |
| SR001 | Pathos AI | Pathos homepage | Pathos integrates advanced AI technology, biological modeling, and the largest collection of oncology patient data to make clinical trials smarter, faster and more precise. |
| SR002 | Pathos AI | About Pathos | Pathos was founded in 2022 by Eric Lefkofsky and Ryan Fukushima. |
| SR003 | Pathos AI | PathOS Platform | We have access to >200 petabytes of multimodal oncology data linked to patient outcomes. |
| SR004 | Pathos AI | Pathos signs strategic agreements with AstraZeneca and Tempus | The agreements include $200 million in data licensing and model development fees to Tempus. |
| SR005 | Tempus AI | Tempus signs expanded strategic agreements with AstraZeneca and Pathos | Tempus’ de-identified oncology data will be used to build the foundation model. |
| SR006 | Tempus AI | Life sciences | 8.5M+ de-identified research records |
| SR007 | Tempus AI | Tempus reports first quarter 2026 results | These forward-looking statements are subject to risks and uncertainties related to: the intended use of Tempus’ products and services; Tempus’ financial performance; ... compliance with new laws, regulations and executive actions, including any evolving regulations in the artificial intelligence space. |
| SR008 | Tempus AI | Our Team | Robyn serves as Tempus’ Chief Privacy Officer. |
| SR009 | Tempus AI | Accelerating a phase 1 oncology trial: The TIME Network’s impact on patient enrollment | The traditional process of identifying, contracting with, and activating clinical trial sites can take many months. |
| SR010 | Pathos AI | Pathos AI doses first patient in phase 1b/2a clinical trial of pocenbrodib | The study is expected to enroll approximately 203 patients with mCRPC. |
| SR011 | Urology Times | Trial launches of CBP/p300 inhibitor in mCRPC | The trial will begin with the phase 1b portion ... and proceed to phase 2a if both acceptable safety and the minimal threshold for efficacy are achieved. |
| SR012 | Pathos AI | Pathos closes $62M Series C financing | This latest round ... was completed at a $600 million post money valuation. |
| SR013 | BioSpace | Pathos expands pipeline with worldwide license of PRMT5 inhibitor | The most common adverse events (grade ≥3), occurring >5% were thrombocytopenia (9.3%), anemia (9.3%), and fatigue (5.8%). |
| SR014 | Pathos AI | Pathos acquires majority stake in DeuterOncology | DO-2 is one of four major portfolio decisions made through Foundry in Q1 2026 alone. |
| SR015 | BioSpace | Pathos AI acquires majority stake in DeuterOncology | peripheral edema rate of just 5% (versus 62-82% for competitors) |
| SR016 | Pathos AI | Pathos AI completes acquisition of Rain Oncology | Rain became a wholly owned subsidiary of Pathos. |
| SR017 | HHS Office for Civil Rights | Guidance regarding methods for de-identification of protected health information | Both methods, even when properly applied, yield de-identified data that retains some risk of identification. |
| SR018 | The ASCO Post | Understanding the legal and ethical challenges AI poses in oncology | Although artificial intelligence tools offer promising advancements in diagnosis and treatment, their legal implications are evolving and uncertain. |
| SR019 | Current Treatment Options in Oncology | Ethical, legal and informed consent challenges for AI-driven therapeutic decision-making in oncology | Bias in AI models remains a major ethical issue. |
| SR020 | National Cancer Institute | Biomarker testing for cancer treatment | Biomarker testing may help you and your doctor choose a cancer treatment for you. |
| SR021 | National Cancer Institute | Cancer statistics | In 2025, an estimated 2,041,910 new cases of cancer will be diagnosed in the United States. |
| SR022 | SEER | Cancer stat facts: all cancer sites | Estimated New Cases in 2026 2,114,850 |
| SR023 | FDA | Direct-to-Consumer Tests | Some direct-to-consumer tests have a lot of scientific and clinical data to support their claims, while other tests do not have as much supporting data. |
| SR024 | FDA | Recalls, Market Withdrawals, & Safety Alerts | The Recalls, Market Withdrawals & Safety Alerts are available on FDA’s website for three years before being archived. |
| SR025 | SEC | Pathos AI SEC submissions | accessionNumber 0001967854-25-000001, 0001967854-24-000001, 0001967854-23-000001 |
| SR026 | SEC | Pathos AI Form D 2025 | Total Offering Amount $399,999,933 |
| SR027 | SEC | Pathos AI Form D 2024 | Total Offering Amount $61,999,979 |
| SR028 | SEC | Pathos AI Form D 2023 | Total Offering Amount $39,999,988 |
| SR029 | Caris Life Sciences | Caris home | 1.0+ Million Cases |
| SR030 | Caris Life Sciences | Caris Life Sciences reports first quarter 2026 financial results | risks related to data security, patient privacy, and compliance with healthcare data protection regulations |
| SR031 | SEC | Caris Life Sciences S-1 | we will avail ourselves of the exemption from the requirement to obtain an attestation and report from our auditors on the assessment of our internal control over financial reporting |
| SR032 | Fierce Biotech | Pathos AI secures $365M series D to fund trials | Pathos hasn’t name-checked the failed liposarcoma therapy in its more recent releases. |
| SR033 | HHS | The Security Rule | The Security Rule |
| SR034 | National Cancer Institute | Can AI Chatbots Correctly Answer Questions about Cancer? | AI chatbots can synthesize medical information, but they are not yet able to consistently generate reliable responses to clinical questions from patients. |
| SR035 | FDA | Software as a Medical Device (SaMD) | Regulators across the globe recognized the need to converge on a common framework and principles for Software as a Medical Device. |
| SV001 | Pathos AI | Pathos secures $365 million Series D financing | Pathos AI ... announced its $365 million Series D financing, bringing its post-money valuation to approximately $1.6 billion. |
| SV002 | Pathos AI | Pathos closes $62M Series C financing | The Series C financing round ... was completed at a $600 million post money valuation. |
| SV003 | Fierce Biotech | Pathos AI secures $365M series D to fund trials | Thursday's release didn't include the identities of the participants in the series D round. |
| SV004 | GenomeWeb | Pathos AI raises $365M series D financing | Pathos claims to be developing the largest oncology foundation model to date. |
| SV005 | Pharmaphorum | Pathos AI’s big $365m series D | P-500 ... has completed phase 1 testing with a mid-stage trial planned. |
| SV006 | Pathos AI | Pathos signs strategic agreements with AstraZeneca and Tempus | The agreements include $200 million in data licensing and model development fees to Tempus. |
| SV007 | Tempus AI | Tempus signs expanded strategic agreements with AstraZeneca and Pathos | The agreements include $200 million in data licensing and model development fees to Tempus. |
| SV008 | SEC | Pathos AI SEC submissions | accessionNumber 0001967854-25-000001, 0001967854-24-000001, 0001967854-23-000001 |
| SV009 | SEC | Pathos AI Form D 2025 | Total Offering Amount $399,999,933 |
| SV010 | SEC | Pathos AI Form D 2024 | Total Offering Amount $61,999,979 |
| SV011 | SEC | Pathos AI Form D 2023 | Total Offering Amount $39,999,988 |
| SV012 | CompaniesMarketCap | Tempus AI market cap | As of May 2026 Tempus AI has a market cap of $8.29 Billion USD. |
| SV013 | CompaniesMarketCap | Recursion Pharmaceuticals market cap | As of May 2026 Recursion Pharmaceuticals has a market cap of $1.59 Billion USD. |
| SV014 | CompaniesMarketCap | Schrödinger market cap | As of May 2026 Schrödinger has a market cap of $0.99 Billion USD. |
| SV015 | CompaniesMarketCap | Relay Therapeutics market cap | As of May 2026 Relay Therapeutics has a market cap of $2.90 Billion USD. |
| SV016 | StockAnalysis | Absci market cap | Absci has a market cap or net worth of $795.12 million as of May 22, 2026. |
| SV017 | StockAnalysis | Recursion Pharmaceuticals market cap and net worth | Recursion Pharmaceuticals has a market cap or net worth of $1.6 billion as of May 22, 2026. |
| SV018 | StockAnalysis | Schrödinger market cap and net worth | Schrödinger has a market cap or net worth of $993.79 million as of May 22, 2026. |
| SV019 | Caris Life Sciences | Caris Life Sciences reports first quarter 2026 financial results | Reported total revenue of $216.2 million |
| SV020 | SEC | Caris Life Sciences S-1 | REGISTRATION STATEMENT UNDER THE SECURITIES ACT OF 1933 |
| SV021 | SEC | Caris Life Sciences 2025 10-K | The aggregate market value ... was approximately $3.87 billion |
| SV022 | Tempus AI | Tempus reports first quarter 2026 results | Revenue of $348.1 million, up 36.1% year-over-year |
| SV023 | Tempus AI | Life sciences | 8.5M+ de-identified research records |
| SV024 | Caris Life Sciences | Caris home | 1.0+ Million Cases |
| SV025 | Recursion | Recursion home | Over the last decade, we have generated and aggregated one of the largest ... datasets in the world — >50 petabytes |
| SV026 | Pathos AI | Pathos homepage | Pathos is an AI-enabled, clinical-stage biotech |
| SV027 | Pathos AI | Pathos launches precision oncology pipeline with license of first phase I program | Pathos obtains worldwide rights from Novo Nordisk for the development of CBP/p300 inhibitor, FT-7051. |
| SV028 | BioSpace | Pathos expands pipeline with worldwide license of PRMT5 inhibitor | Out of 16 patients with high-grade glioma ... two confirmed complete responses were observed. |
| SV029 | Marketscreener | Caris Life Sciences quote page | Caris Life Sciences Inc. is a patient-centric, next-generation artificial intelligence tech bio company and precision medicine provider. |
| SV030 | Tempus AI | Tempus Oncology | 6.5K+ Oncologists rely on Tempus |
| SV031 | Pathos AI | Pathos AI doses first patient in phase 1b/2a clinical trial of pocenbrodib | The study is expected to enroll approximately 203 patients with mCRPC. |