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

Noetik is building foundation models trained on multimodal patient data to solve the 95% failure rate of cancer trials by redefining patient selection, not drug design. The company argues that most oncology drugs fail not due to poor pharmacology but because they're tested on poorly characterized patient populations, often modeled on artificial cell lines that don't reflect human biology. Their approach uses transformer models trained on real tumor samples to identify functional disease subtypes and match drugs to patients based on underlying biology.

You'll understand how foundation models applied to real patient biology—not synthetic cell lines—could fix the broken paradigm of cancer drug development.

1. 01Cancer drug failures are primarily due to flawed patient selection, not ineffective molecules, because preclinical models like immortalized cell lines don't reflect real human tumor biology.
2. 02Noetik trains transformers on multimodal data from real human tumor samples to discover hidden disease subtypes and predict which patients will respond to specific therapies.
3. 03The company's models operate at the level of functional tissue rather than individual cells or genes, enabling a more accurate, patient-level prediction of treatment response—similar to how neural networks abstract neurons in computational neuroscience.