Study On Computational Techniques For Flexible Protein Docking

Complexes play a critical role in signal transduction and material transportationof biological function. As the complex function decided by its’3D structure, so thestructure prediction can help us to know its’ function when the experimental structureis unknown. But the prediction of complex structure is still a huge challenge at present.As the proteins structure varied when docking, taking the fexibility into considerationbecomes extraordinary difcult for the structure prediction. This article studied arecognition of fexible region and the application of the probability graph model forfexible docking.First, the fexible region recognition of G-Protein Couple Receptor is studied.The G-Protein Couple Receptor plays an extremely important role in drug design.The transmembrane is the main feature of the G-Protein Couple Receptor, and themembrane usually has fexibility. Because of the membrane fexibility, the prediction ofthe G-Protein Couple Receptor becomes more difcult. So, it is important to recognizeand predict the fexible region of the membrane. We clustered the membrane intodiferent groups according to the sequence similarity, and used von Misses continuousmodel to train the fexible distortion angles. Our trained model only needs to sampleffteen times to get at least one sample near native distortion angle. Improving theprecision of membrane fexible region recognition is really helpful to fexible regionscomputation for G-Protein Couple Receptor structure prediction and docking.Second, the backbone fexible is supplemented into protein-protein docking pro-tocol based on Rosetta Dock for improving the docking precision. Improving thecomplex prediction relies not only to improve the sampling ability during the dockingprocedure but also to advance the capability of energy scoring function in picking thedocking candidates. Incorporation of fexibility expanses of decoys energy range that inevitably aggravates the difculty of the energy function selection. This article usedgraph probability model called as matrix Bingham-von Misses-Fisher to model thefexible docking decoys’ residue interaction graph. After training, the model can beused to compute the probability of a decoy. As if can quantify the biological impactinto structure, the residue interaction graph could be the accordance of candidatesselection. Our model is a useful supplement for energy function in ranking fexibledocking decoys, especially in cases where the energy function failed to pick out mostpromising candidate structures.