Generable

A Bayesian computational oncology platform that predicts individual disease progression under different therapeutic choices to reduce treatment failure in trials and clinical care. It connects low-level measurements such as drug concentrations, tumor size, and ctDNA to outcomes including PFS/OS and adverse events to inform dosing, biomarker evaluation, and trial design. It sells to biotech and pharmaceutical companies and hospitals. It is positioned as explicitly coded, explainable models using Bayesian inference with HMC/NUTS and variational methods that provide calibrated probabilistic predictions and work in small-data settings using prior knowledge.

Key features

• Predict disease progression on or off treatment
• Compare treatment vs competitor or standard of care
• Determine best dose for treatment or person
• Evaluate biomarker predictiveness
• Differentiate responders from non-responders
• Design Bayesian clinical trials
• Connect drug concentration to PFS/OS outcomes
• Connect tumor size to clinical outcomes
• Connect ctDNA to adverse events
• Explainable models defined explicitly through code
• Bayesian inference with HMC and NUTS
• Pathfinder Variational Inference for large models
• Calibrated probabilistic predictions
• Small-data modeling using prior knowledge