The Improvement And Application Of Protein Secondary Structure Prediction Method PSIPRED

Understanding the protein-protein interactions and how proteins to exert its biological function are very important for biology,medicine and pharmacy,and it is necessary to know the three-dimensional structure of proteins,thus predicting the three-dimensional structure of proteins is imperative.However,predicting protein structure directly from primary sequence is almost impossible.A large number of studies have shown that if the secondary structure of the sequence is known,to predict its three-dimensional structure will become feasible,therefore,protein secondary structure prediction is the first step in protein structure prediction.Firstly,this paper studied how to build a more reliable prediction model of protein secondary structure,including designing effective feature extraction algorithm and classification algorithm.Then,this paper applied protein secondary structure information to bacterial virulent proteins prediction.The main content of the research work and innovation points was summarized as follows:A new feature extraction algorithm is proposed in this paper to improve the secondary structure prediction accuracy,which is Combined dihedral Angles with the position specific scoring matrices as input profile for the classification model of two-layered neural network,and every amino acid residue in the center of the sliding window was represented by a 22*w(w represents the window length)dimensional numerical characteristic vector.The proposed method was applied to the two most frequently used datasets CB513 and RS126,the results of cross validations was compared with that of nine existing methods.Moreover,a comparison with 10 online secondary structure prediction servers was made with an independent test on CASP9 targets.Due to bacterial virulent proteins can be a drug target bacteria or vaccine candidate,developing a reliable prediction model of bacterial virulent proteins prediction is helpful for finding novel drug targets,vaccine candidates,and understanding virulence mechanisms in pathogens.In this paper,the predicted secondary structure-based features were firstly applied to the bacterial virulent proteins prediction.A combination of feature extraction methods with an ensemble of SVMs for bacterial virulent proteins prediction was developed.Sequence-based features(dipeptide composition and Chou's Pseudo Amino Acid composition),predicted secondary structure-based features were used to train four base classifiers respectively.The results on the three most frequently used datasets show that the approach proposed in this paper has gained higher overall classification accuracy.