Home > About > Events & Seminars > A physics-based framework for knowledge-based protein models: Directly determining physical potentials from a model protein databank

A physics-based framework for knowledge-based protein models: Directly determining physical potentials from a model protein databank

Monday, September 17, 2PM – 3PM
ACE 6.304

William Noid

Low resolution coarse-grained (CG) models play a central role in many areas of computational protein science. Since the seminal work of Tanaka and Scheraga, many groups have employed structural statistics extracted from the protein databank (PDB) to parameterize "knowledge-based" potential functions for CG protein models. However, it has proven challenging to provide a first principles theory for knowledge-based potentials determined from the PDB. It is well established that conventional knowledge-based approaches do not correctly treat the correlations between coupled interactions in protein structures. Furthermore, it has not been clear how conventional statistical thermodynamics can be applied to analyze structural correlations for different proteins. The present talk will discuss our recent work addressing these two challenges. We have derived the first generalization of the Yvon-Born-Green integral equation theory for complex molecules, such as proteins, with flexible internal degrees of freedom. This theory determines a variationally optimal potential function directly from structures. In addition, we have developed an extended ensemble formalism for developing transferable potentials from structural correlations for multiple proteins. In combination, these two advances provide a physics-based framework for determining transferable potentials from a databank of structures for multiple proteins. Additionally, recent work has shown intriguing connections between this approach and information theoretic approaches to multiscale modeling.

Hosted by Dr. Dmitrii Makarov