Modeling of Biomolecular Systems Interactions, Dynamics, and Allostery: Bridging Experiments and Computations - September 10-14, 2014, Istanbul, Turkey

Modeling of Biomolecular Systems Interactions, Dynamics, and Allostery Session II Abstracts

Soman Induced Conformational Changes of Human Acetylcholine Esterase Sebnem Essiz Gokhan 1 , Brian Bennion 2 , Edmond Y. Lau 2 , Felice C. Lightstone 2 . 1 Kadir Has University, Istanbul, Turkey, 2 Lawrence Livermore National Lab, Livermore, CA, USA. Permanent inhibition of acetylcholine esterase, AChE, results in “runaway” neurotransmission leading to cognitive deficiencies, seizures, paralysis, and eventually death depending on the exposure. We present data from quantum mechanics/molecular mechanics (QM/MM) and 100 ns (MD) simulations of the apo and soman-adducted forms of hAChE to investigate the effects on the dynamics and protein structure when the catalytic Serine 203 is phosphonylated. By using correlation and principal component analysis of MD trajectories, we identified the allosteric sites in addition to segments of the protein, which are loosing flexibility due to the presence of soman in the binding pocket. The altered motions and resulting structures provide for alternative pathways into and out of the enzyme active site through the side-door in the soman-adducted protein.

Evolution Teaches Predicting Protein Interactions from Sequence Burkhard Rost . Technical University of Munich, Garching, Germany.

The physical protein-protein interaction (PPI) between two proteins A and B can be predicted from sequence alone. However, methods perform poorly on this difficult task when both proteins A and B have not been in the training set. Tobias Hamp in our group has developed a new approach that improves significantly over state-of-the-art methods. We machine learned highly reliable human PPIs from the Hippie resource through a new profile-kernel based SVM. This use of evolutionary information in combination with predicted sub-cellular localization raises precision even for low recall levels (most reliable predicted few interactions). A new rigorous way to reduce PPI redundancy reveals that only a fraction of the available PPIs is needed to build more accurate classifiers.

19