Generable

Generable is a New York-based scientific computing company that applies Bayesian statistical methods to problems in drug development and clinical decision-making. Founded by developers of the Stan probabilistic programming language, the company positions itself as a provider of open-source model development, optimization and training. Its research connects low-level measurements — such as drug concentrations, tumor size and ctDNA — to clinically meaningful outcomes including progression-free and overall survival and adverse events, and addresses questions such as Bayesian trial design, biomarker predictiveness, responder characterization, disease progression, comparative treatment probability and dose selection.

The company offers several model families. Joint outcome-biomarker models link hazard models with submodels for individual biomarkers and comorbidities, and are used in Phase II and Phase III to relate individual biomarker evolution to event risk and to quantify biomarker impact on clinical endpoints; Generable applies non-parametric and semi-mechanistic biomarker models, hierarchical models with external controls, causal estimands and post-stratification. Complex PK/PD models use ordinary differential equations to represent drug distribution in the body and are applied in preclinical and Phase I work to establish safe dosing regimens while detecting early efficacy signals, inferring both population- and individual-level parameters. Biomarker models use non-parametric Bayesian methods such as Gaussian processes and BART to model wide data where clustering and dimensionality reduction are considered unlikely to surface clinical relationships.

Founding story

Generable was started by developers of the Stan probabilistic programming language. Co-founders include Eric Novik (CEO), an applied statistician and entrepreneur who mentors in Columbia University's Statistics Department and is adjunct faculty of statistics at NYU; Jacqueline Buros (CSO), a biostatistician and bioinformatician previously with the TIMI Study Group at Harvard Medical School, Alzheimer's disease genetics at Boston University, and lead biostatistician at the Institute for Next Generation Health Care at Mount Sinai; and Daniel Lee (Co-Founder Emeritus), an early Stan contributor who developed Stan's ODE subsystem and built PK/PD models including one for an FDA-approved drug.

Business model

Generable works with pharmaceutical and biotechnology companies, providing model development, optimization and training around Bayesian and probabilistic methods applied to clinical development questions.

Traction

The company states it works with pharmaceutical and biotechnology companies on trial design, biomarker and dosing questions, and publishes blog analyses of clinical trial readouts, including predictions for the LEVEL trial of oral levosimendan in PH-HFpEF and an evaluation of predictive performance for the ACACIA-HCM hypertrophic cardiomyopathy trial.