Featurization of Protein-Protein Interfaces for Machine Learning in Large Molecule Drug Discovery

Time: 2:00 pm
day: Day One


  • Atomic-level characterization of protein interfaces can be computed from all experimental, and improved modelled, structures to generate a large dataset (4×10^6) suitable for machine learning input
  • We created PiEx, a fully automated platform for protein:protein interface analysis, that allows navigation and exploration via interactive visuals of the large data
  • PiEx generates novel insights from native and crystal interfaces, as exemplified in B-cell and T-cell receptor studies