Using Evolutionary Conservation and Rigidity Analysis for Protein Docking Refinement

October 25, 2013
11:00a - 12:00n
Halligan 209
Speaker: Nurit Haspel, Dept of Computer Science, UMass Boston


Proteins often bind to other proteins to create complexes that act like molecular machines. These complexes play a central role in nearly every process in the cell. Since the three dimensional structure and the functionality of proteins are closely related to each other, detection of protein complexes and their structures is crucial for understanding the role of protein complexes in the basic biology of organisms. Predicting the structure of a complex formed by assembling multiple chains is a difficult problem to solve experimentally. Computational methods can become very useful and provide researchers with a good starting point for analyzing protein complexes. Computational docking methods are typically made of two stages: The search stage uses structural and geometric techniques to detect native-like configurations of the complex, and the ranking stage uses a scoring function made of physico-chemical and geometric filters to estimate the binding affinity and rank computed structures according to energetic criteria.

Computational docking methods are still not accurate. The energetic difference between the native structure and other non-native complexes may be small and the scoring function used by docking methods is often not sensitive enough to detect it. Additionally, the correct binding site is not always known experimentally and docking methods may miss the correct binding site completely. As a result, low-energy structures produced by docking programs often disagree with NMR data .

I will discuss my recent work at developing a multimeric docking refinement scheme that uses a scoring function based on a tight coupling between evolutionary conservation, geometry and physico- chemical interactions. I will discuss current efforts of combining evolutionary conservation data with graph-based rigidity analysis and machine learning to detect critical residues in proteins, which may aid in detecting binding interfaces and model conformational changes in proteins.