TITLE: Predictive Modeling of Protein Interactions and Rational Design of High Affinity Binding

SPEAKER: Dr. Brian Pierce

TIME: Thursday February 4 at 11 AM

PLACE: George P. Williams, Jr. Lecture Hall, (Olin 101)

Protein-protein interactions play a key role in many biological processes, and a predictive understanding of their behavior would greatly improve our ability to treat and cure disease. By utilizing the wealth of available structural data, we have developed models to predict structures of protein complexes (also known as protein docking) and to design protein interactions for high affinity and specificity. The protein docking programs include M-ZDOCK, for prediction of symmetric multimer assemblies from structures of individual proteins, and ZRANK, which quickly and effectively rescores initial docking predictions. The latter work in scoring was extended to perform structure-based design of a T cell receptor (TCR) for improved binding to a peptide/MHC complex; this immunological interaction leads to cellular immune response against tumors and viral antigens. Based on initial measured data for this system, we developed a function to model affinity changes of point mutations, incorporating structural predictions from the program Rosetta. This led to several point mutations with significant affinity improvement over wild-type, and when combined, binding affinity was improved by 100 times. We found striking non-additive effects when combining point mutations in the TCR and analyzed them according to binding kinetic rates. We then utilized this algorithm to analyze other protein complexes for designability, and determined point mutations on the CD4 protein for improved HIV gp120 binding. Future applications of this work include analysis and design of an anti-melanoma TCR, improved scoring of protein complex predictions, and de novo protein interface design.