🔬 Training Transformers to solve 95% failure rate of Cancer Trials — Ron Alfa & Daniel Bear, Noetik

Description

TL;DR: 95% of cancer treatments fail to pass clinical trials, but it may be a matching problem — if we better understood what patients have which tumors which will respond to which treatments, success rates improve dramatically and millions of lives can be saved — with the treatments we ALREADY have! GSK recently signed a $50M deal for their technology that also includes an (undisclosed) long-term licensing deals for Noetik’s models. Most big AI plays in BioTech have focused on discovery, and usually result in an in-house development effort (meaning tools companies usually become drug companies). This deal stands out in that it is a software licensing deal, and represents a commitment to a *platform* rather than a *drug*. With attention on other software tools for drug development (see the [Boltz episode](https://www.latent.space/p/boltz) and Isomorphic for example), it is starting to look like the appetite of Pharma for biotech tools has finally started to grow. Why the sudden interest? Timestamps: (0:00) The challenges of starting a biotech lab and generating data from scratch. (0:55) Introduction of Ron Alfa and Dan Bear from Noetik. (4:09) The complexity of cancer: Why "curing cancer" is a misleading concept and the need for new, multimodal data. (8:24) Identifying therapeutically relevant cancer subtypes to improve clinical trial success rates. (11:27) The importance of intentional, high-quality data generation in AI biotech. (17:09) Lessons learned from Recursion Pharmaceuticals regarding batch effects and data design. (20:14) Introduction to Noetik's core data modalities: Pathology (H&E), spatial transcriptomics, and genomic alterations. (30:15) The philosophy of self-supervised learning and avoiding bias from electronic health records. (36:01) Translating latent space embeddings and patient clusters into actionable insights for pharma. (41:40) Using PerturbMap and in-vivo mouse models to validate human AI predictions. (53:38) Technical deep dive: The Tario transformer-based model and auto-regressive training objectives. (1:00:26) The GSK partnership: Licensing OctoVC and the shift toward platform-based biotech deals. (1:13:55) Advice for small biotech AI startups: Scaling, data conviction, and lessons from scientific history.