High Accuracy Protein Structure Prediction by Global Optimization
Dec 11, Tue 2007
11:00 am, MRB 200 Conference Room
School of Computational Sciences, Korea Institute for Advanced Study
We propose a new high-accuracy protein structure modeling method based on straightforward and rigorous application of a global optimization to the three consecutive layers of modeling: multiple alignment, chain building, and side-chain modeling. We applied the conformational space annealing method to a consistency based score function for multiple alignment. For chain building and side-chain modeling, we optimized the MODELLER energy function and a SCWRL-like energy function using a rotamer library constructed specifically for a given target sequence. The method was applied to CASP7 targets in a blind fashion. Assessment of the prediction, conducted after the native experimental structures are released, demonstrates that significant improvements over existing methods are achieved in backbones as well as in side-chains for TBM (template-based modeling) and high-accuracy TBM targets. For most high-accur acy TBM targets (17/26), the predicted model was more accurate than the model one can construct from the best template in a posteriori fashion. It appears that the proposed method can extract relevant information out of multiple templates. Advantages as well as limitations of the method will be discussed.