Somerra.ai combines programmable organoids with machine learning to create a Virtual Organ that predicts how human tissue responds to untested drugs and conditions. Each organoid experiment trained on patient-derived stem cells improves the model across donors, drugs, doses, and schedules, enabling the system to select the next most informative test. This iterative approach aims to reduce the years, capital, and human risk associated with traditional clinical trials.
Funding
Funding not disclosed
Founders
Product
Problem
Clinical trials require years of time, substantial capital, and expose human participants to risk in order to determine whether a drug is safe, effective, and for whom. The knowledge gained from each trial is largely siloed within a single program, so learning is not cumulative across drugs, doses, or patient populations, making drug development slow and inefficient.
Solution
Somerra.ai builds a Virtual Organ—a predictive model trained on data from programmable organoids, which are living three-dimensional human tissues derived from patient stem cells using gene circuits pioneered at MIT. Each organoid experiment tests drugs, doses, and schedules across multiple donorsainer, and the resulting data continuously trains the Virtual Organ. The model then predicts how untested conditions will behave and selects which experiment to run next, closing a loop that accelerates discovery. This platform enables researchers to explore far more of the therapeutic space than traditional trial methods allow, while reducing reliance on costly, time-consuming human studies that start from scratch each time.
Target Audience
Somerra.ai serves pharmaceutical companies, biotech firms, and clinical research organizations seeking to de-risk drug development by replacing or augmenting conventional preclinical and clinical testing with predictive tissue models.
Features
- Programmable organoids created from patient-derived stem cells using synthetic-biology gene circuits, enabling reproducible and customizable human tissue models
- Virtual Organ platform that aggregates experimental data across donors, drugs, doses, and schedules to make predictions about untested conditions
- Active learning loop where the model recommends the next experiment to maximize information gain and iteration speed
- Machine-learning infrastructure designed and scaled by co-founder who previously built ML systems for protein design at Generate:Biomedicines
- End-to-end workflow that transforms raw organoid data into predictive insights for drug development decisions