Research On Several Basic Issues In Prediction Of Biomolecular Complex Structure

Biomolecules such as proteins and RNA are important substances that make up organisms,and they participate in most physiological processes of organisms by interaction with each other.In order to understand the functions and mechanisms of these biomolecules in physiological processes,the structures of biomolecules are necessary.For the current experimental techniques,it is easier to obtain the sequences than to obtain structure of protein and RNA,and it is easier to obtain the structures of the monomers than to obtain the structures of complexes.Therefore,for further research,computational methods for unknown biomolecules structures from existing structural and sequence information have been developed.Based on computational methods,this thesis addresses three issues:1.Development of scoring functions of the protein-RNA complex structure.At present,the prediction of biomolecule complex structures includes two steps:sampling and scoring.And the scoring functions can be divided into physics-based scoring functions and knowledge-based scoring functions.There are two factors that affect knowledge-based scoring functions:statistical objects and reference state.In this thesis we propose two novel knowledge-based scoring function with different statistical objects.a)Score_?is a knowledge-based scoring function which uses the stacking pairs between residue and base as statistical object.This scoring function adds a stacking item to the existing scoring functions,so that the scoring functions is easier to get the near-native structures.b)Different from the usual distance/angle-dependent scoring function,3dRPC-Score is a scoring function which uses conformations of the residue-nucleotide pairs as statistical object.And on different testing sets,3dRPC-Score gives a consistent performance.2.Research on the bound-state tertiary structure characteristics of intrinsically disordered protein with the co-evolution information.The intrinsically disordered protein(IDP)is a kind of proteins widely present in living organisms,which participates in various physiological processes,such as molecular and cellular regulation and signaling.The sequences of IDPs consists of a large proportion of hydrophilic residues and charged residues.As a result,IDPs are hard to form a stable folded structure by themselves in solution.And the structures of IDPs are difficult to observed by experimental methods.Instead of stable folded structures,most of IDPs have a conformational ensemble,and transform in several conformations dynamically.Besides,some IDPs fold into a stable structure when bound to their partners.Therefore,it's important to determine the bound-state conformation of IDP,when predicting the structure of the IDP-related complex.In this thesis,we extract the coevolutionary residue pairs(co-pair)map from the multiple sequence alignment(MSA)of IDP by direct coupling analysis(DCA)and compare with the contact map of the bound-state structure of IDP.As a result,the co-pair map can be used to predict the contacts of bound-state IDP,which means that the co-evolution information of IDP encode the tertiary structure of bound-state conformation of IDP.On one hand,this indicates that IDP satisfy the sequence-structure relationship of protein,on the other hand,the co-evolution information of IDP is helpful to construct the bound-state IDP structure.3.Validation and analysis of the predicted biomolecular complex structure with molecular dynamics simulation.After obtaining the predicted biomolecular complex structure by computational method,how to verify whether it is a reasonable structure,and how to analyze its interaction mechanism of the complex are also difficult problems in the prediction of biomolecular complex structure.In this thesis,we try to solve these problems by molecular dynamics simulation.In this way,the anti-HIV-1 virus drug design which target the cyclin-dependent kinase(CDK)was carried out together with co-workers,and some results have been achieved.