Research Of Information Entropy Based Protein Strucutre Prediction Methods

Proteins are macromolecules in organisms and the principal undertakers in life activities.Genetic code reveals the transfer machanisms from DNA to amino acids.However,how fold in specific spatial structures of proteins remains a mystery.Only protein with specific spatial structures can conduct specific biological functions.In this case,there is significantly important theoretical value to study protein spacial structure,as for understanding of the relationship between amino acid sequences and spatial structures.Nowadays,with the rapid development of sequencing technology,the development of technology of experimentally structural determination is relatively slow.The situation results in the exponential increasing of the gap between the number of knew amino acid sequences and structures.Driven by the theoretical value and the practical significance,it is one of the important research topics in the field of bioinformatics to predict the three-dimensional structure of proteins from amino acid sequences by computers and algorithms.De novo prediction is an important method for protein structure prediction According to Anfinsen's dogma,the method is guided by energy function and employs optimization algorithms to sample comformational space,with expect to search and select near-native conformations.However,the comformational space of proteins is too vast and the energy landscape is extremely rough so that de novo prediction is one of the most challenging tasks in protein structure prediction.As for the problem,the main work of the thesis is summarized as follows:(1)As for protein structure prediction,contact map based residue-pair distances restrained protein structure prediction algorithm is proposed in the thesis.In the exploration stage of the algorithm,residue-pair distances-biased mutation and selection strategies are designed through the discretization of residue-pair distances for the purpose of maintaining the diversity of the population.In the exploitation stage of the algorithm,energy function and contact map based scoring index are selected according to a probability,which can enhance the exploration ability of conformational space.Finally,the algorithm is tested on diverse target proteins to verify its effectiveness.(2)As for the problem of multistage strategies switch with fixed computational cost in the field of protein structure prediction,a protein structure prediction algorithm guided by stage switch is proposed.In the exploration stage of the population evolution,an information entropy based stage switch factor is constructed through the distribution of dihedral angle to reflect the diversity of the population.The adaptive switch is implemented by employing the change of the factor.Meanwhile,accumulated dihedral angle distribution in the exploration stage is used to establish a scoring index in the exploitation stage,which is combined with energy function to further sample the conformational space.The experimental results indicate that the algorithm can predict protein three-dimentional models with high accuracy.