Targeting Protein-Ligand Neosurfaces with a Generalizable Deep Learning Tool

Watch a research presentation exploring a novel computational strategy for designing proteins that target neosurfaces formed by protein-ligand complexes. Discover how geometric deep learning and learned molecular surface representations are leveraged to develop and validate protein binders against three drug-bound protein complexes: Bcl2-venetoclax, DB3-progesterone, and PDF1-actinonin. Learn about the high affinities and specificities demonstrated through mutational and structural characterization, and understand how surface fingerprints trained on proteins can be successfully applied to small molecule-induced neosurfaces. Explore the potential applications of these chemically induced protein interactions in expanding sensing capabilities and developing new synthetic pathways for drug-controlled cell-based therapies.