AI x Science Seminar: Pranam Chatterjee

The Chatterjee Lab at the University of Pennsylvania develops generative algorithms to design functional molecules directly from sequence. Our work begins with a question: can short peptides be de novo designed to bind undruggable targets like disordered oncogenic fusions, neurodegenerative aggregates, and viral phosphoproteins? With our language models including PepPrCLIP and PepMLM, we’ve shown that the answer is yes. We’ve extended our models to enabling targeting of motifs (moPPIt), post-translational modifications (PTM-Mamba), fusion-driven disease mechanisms (SOAPIA), and even toxic heavy metals (Metalorian). Beyond binding, we’ve built theoretical frameworks for discrete generation like PepTune and MOG-DFM that optimizes sequences across multiple therapeutic properties such as target affinity, specificity, solubility, and permeability. Most recently, we unify molecular design with dynamics and state modeling using Schrödinger bridge formulations that learn not only relevant biological endpoints (BranchSBM) but also the trajectories between them (EntangledSBM), spanning protein folding, molecular transitions, and cell-state change. In this talk, I’ll share how these advances form a unified framework for programming biology through sequence models trained on and guided by biological constraints.

Presented by
Penn AI, Innovation in Data Engineering and Science (IDEAS),
and the Data Driven Discovery Initiative (DDDI)