The Machine Learning team develops predictive models of biological function to help experimental biologists design synthetic biomolecules with desired characteristics. Working closely with the wet lab teams, we connect computational design with laboratory testing to improve study design, anticipate outcomes, and prioritize promising candidates. Results from in vitro and in vivo experiments inform successive rounds of model refinement and focus on biomolecule optimization.

The team combines generative modeling, iterative optimization, hypothesis-driven active learning, and causal modeling to explore biological possibilities and design informative experiments. We also investigate how human expertise and knowledge encoded in large language models can help researchers formulate scientific hypotheses and design experiments to test them.