Outer Biosciences now feeds live human skin—harvested from discarded surgical tissue and kept viable for more than a month—into machine-learning models. The startup hopes to make drug-discovery and dermatology predictions far more reliable. Its two-year pipeline moves tissue from operating rooms to a lab within hours, a speed the company says preserves the cells’ natural behavior.
Why the old data-sets fall short
For most of biotech’s recent history, researchers have relied on animal models or cultured cell lines to train computational tools. Those systems are cheap and easy to handle, but they rarely capture the full complexity of human skin—a layered organ with immune cells, nerve endings, microbiome interactions and a dynamic response to hormones and stress. The mismatch forces drug developers to run multiple rounds of animal testing and human trials, inflating costs and extending timelines.
Outer Biosciences positions itself as a bridge across that gap. By using tissue that has never been altered in a dish, the company believes its AI can learn patterns invisible to synthetic models. “Living data” is the phrase the team uses to describe the difference between a static snapshot of a cell line and a continuously interacting tissue slice that still communicates with its environment.
From the operating room to the algorithm
The logistical challenge matches the scientific one. Outer Biosciences sources skin that would otherwise be discarded during procedures such as abdominoplasties or mastectomies. Nonprofit and commercial biobanks, operating under Institutional Review Board (IRB) oversight, document donor consent for research use. Once a hospital flags a suitable specimen, the company’s courier network picks it up, transports it in temperature-controlled containers, and delivers it to a processing facility within a few hours.
In the lab, the skin sits in a perfusion system that mimics blood flow, keeping the cells alive for up to 30 days. During that window, the team captures high-resolution imaging, transcriptomics and proteomics data at multiple time points. The multi-modal dataset feeds deep-learning pipelines that predict how the tissue will react to chemical compounds, UV exposure, or inflammatory triggers.
The people pulling the lever
Co-founder and CEO Michael Polansky brings a mathematics background from Harvard and a track record of scaling complex operations, from venture-capital portfolios to global concert tours. Chief Scientist Kyung-Jin Jang, a skin-biology specialist, directs the experimental side, while CTO Chris Hinojosa builds the data-engineering platform that stitches together imaging and molecular streams. The board includes pop-culture icon Lady Gaga, who also chairs the cosmetics brand Haus Labs, a venture that shares research resources with Outer Biosciences.
These cross-industry ties give the startup both the capital to build a niche logistics network and a ready market for early applications. Haus Labs, for example, could test new formulations on the live-tissue platform before moving to human volunteers, potentially shaving weeks off the typical product-development cycle.
What’s at stake
If the approach lives up to its promise, pharmaceutical companies could cut the number of animal studies required for early-stage screening, reducing both cost and ethical concerns. Dermatology clinics might gain AI tools that predict how a patient’s skin will respond to a prescription cream, enabling truly personalized treatment plans. The broader AI community could also benefit from a richer, more realistic training set that pushes the limits of predictive biology.
The model raises questions about scalability and privacy. Collecting tissue from surgery depends on hospitals and donors willing to participate, and the supply chain may bottleneck as demand rises. While biobanks operate under IRB oversight, detailed molecular profiles could, in theory, be linked back to individuals if data-handling practices slip. Critics argue that even with consent, the commercial exploitation of discarded tissue skirts a thin ethical line.
Counter-point: Do organoids still have a role?
开发类器官（在实验室中培养的、用于模拟器官功能的细胞三维集群）的科学家们警告称，不要完全摒弃这些模型。类器官可以从单个患者的诱导多能干细胞中产生，提供了一个无需新鲜手术组织的个性化平台。它们还允许研究人员进行长达数月的发育过程研究，这是为期 30 天的皮肤切片无法做到的。Outer Biosciences 承认，其平台并非万能的替代品，而是在皮肤原生结构最为关键的特定问题上，提供高保真数据的补充来源。
未来之路
Outer Biosciences 目前正与几家生物技术合作伙伴进行试点研究，通过已知的药物结果来测试其模型的预测能力。仅在内部简报中分享的早期结果表明，其与临床试验数据的相关性高于同类基于类器官的预测。该公司计划扩大其组织收集范围，以涵盖其他器官类型，尽管每种类型都会带来各自的物流障碍。
监管机构尚未针对基于活体人类组织训练的 AI 模型发布具体指导意见，但 FDA 针对 AI 驱动医疗设备的新兴框架可能会与 Outer 的工作产生交集。这家初创公司表示，正在准备既符合数据隐私标准，又符合该机构对算法透明度预期的文档。
核心要点
Outer Biosciences 已将一项物流壮举——在短短几小时内将新鲜人类皮肤从手术室转移到实验室——转化为一个数据引擎，这可能使 AI 对药物和皮肤健康的预测更加可靠。该项目的成功将取决于其维持组织供应管道持续运转、保护捐赠者隐私，并证明“活体数据”确实优于现有合成模型的能力。如果成功，其连锁反应可能会波及制药、化妆品和临床皮肤病学领域，重塑我们训练机器理解人体的方式。