A Study On Specific Protein-Peptide Interaction Prediction

Specific binding of Antibodies with the corresponding antigen specific binding can lead to physiological or pathological effects,and then bring about the immune response.Specific surface ligands binding to receptors can activate specific pathways.In addition,interaction between different proteins within the cell can trigger some cellular regulation.In this paper,we focus on the prediction of protein-peptide interaction.The first one focuses on prediction of specific binding of 14-3-3 protein and phosphorylated peptides,and then,we use sequence similarity and structural similarity to predict the binding of MHC molecules and peptides.The 14-3-3 proteins are a highly conserved family of homodimeric and heterodimeric molecules,expressed in all eukaryotic cells.Between the seven isoforms of this family,14-3-3σ is the only isoform directly linked to cancer in epithelial cells.We present the first computational method for identifying peptide motifs binding to 14-3-3σ isoform.In this paper,for each 14-3-3 isoform,we have 1,000 peptide motifs with experimental binding affinity values,and we have to identify the binding value of all the 16,000 peptides.First,we select nine physicochemical properties of amino acids to describe each peptide motif.We also use auto-cross covariance to extract correlative properties of amino acids in any two positions.Then,we propose a similarity-based Undersampling method and a SMOTE-based weighted Oversampling method to deal with the unimbalanced dataset.Finally,we consider locally weighted regression and elastic net to predict affinity values of peptide motifs.MHC molecule plays a key role in immunology,and the molecule binding reaction with peptide is an important prerequisite for T cell immunity induced.In this paper,we propose a novel prediction method for predicting MHC II binding peptides.First,we calculate sequence similarity and structural similarity between different MHC molecules.Then,we re-order the pseudo sequences according to descending similarity,and use a weight calculation formula to calculate the new pocket profiles.Finally,we use three scoring functions to predict binding cores,and evaluate the accuracy of prediction to judge performance of each scoring function.In the experiments,we set a parameter α in the weight formula.By changing α value,we can observe different performances of each scoring functions.Then,we compare our method with the best function to some popular prediction methods,and ultimately find that our method outperforms in identifying binding cores of HLA-DR molecule.