The Modeling And Prediction Of Protein–peptide Affinity And Specificity

Modeling and predicting the protein/peptide binding affinity and specificity is of great importance to the Physiological and pathological process in cells. By taking into the prediction and modeling of the binding affinity and specificity in protein/peptide complex, we offer new thoughts to the computer-aided peptide design. This research mainly contains the following three parts:1. Peptides with antihypertensive potency have long been attractive to the medical and food communities. However, serving as ideal functional foods, peptides should have highly therapeutic activity and a good taste. The T scale coupled with partial least squares(PLS) regression is used to perform quantitative sequence-activity relationship(QSAR) between the angiotensin-converting enzyme(ACE) inhibition and bitterness of short peptides. The results reveal a significant positive correlation between the ACE inhibition and bitterness of dipeptides, but this correlation is quite modest for tripeptides and, particularly, tetrapeptides. Moreover, QM/MM scheme is designed to analyze the complex structure of ACE protein with inhibitory peptides. Results unralve that peptides with 4 amino acids are sufficient to have efficient binding to ACE, and more additional residues can not bring with substantial enhance in their ACE-binding affinity and, thus,antihypertensive capability. Above all, It suggest that the tripeptides and tetrapeptides could be considered as candidates for seeking ideal potential functional food additives with both high antihypertensive activity and low bitterness.2. Gaussian process(GP) is constructed on the basis of Bayesian probabilistic inference, it is a good machine learning method for nonlinear classification and regression, but has only very limited applications in the field of computer-aided vaccine design(CAVD)and immunoinformatics. In this work, GP is introduced into peptide statistical modeling to predict the binding affinities of over 7000 immunodominant peptide epitopes to six types of human major histocompatibility complex(MHC)proteins. In this study, the sequence patterns of different peptides are characterized quantitatively and the resulting variables are then correlated with the experimentally measured affinities between MHC proteins and peptide epitopes, by using GP approach.We compared modeling methods SVM(support) vector machineand PLS(partial least square) with GP to examine the statistical performance in the fitting ability, predictivepower and generalization capability. The results suggest that GP could be a new and effective tool for the modeling and prediction of MHC–peptide interactions and would be promising in the new areas of computational vaccinology.3. Domain-peptide recognition and interaction are fundamentally important for eukaryotic signaling and regulatory networks. It is thus essential to quantitatively infer the binding stability and specificity of such interaction based upon large-scale but low-accurate complex structure models which could be readily obtained from sophisticated molecular modeling procedure. In the present study, a new method is described for the fast and reliable prediction of domain–peptide binding affinity with coarse-grained structure models. This method is designed to tolerate strong random noises involved in domain-peptide complex structures and uses statistical modeling approach to eliminate systematic bias associated with a group of investigated samples.As a paradigm, this method was employed to model and predict the binding behavior of various peptides to four evolutionarily unrelated peptide-recognition domains(PRDs),i.e. human amph SH3, human nherf PDZ, yeast syh GYF and yeast bmh 14-3-3, and moreover, we explored the molecular mechanism and biological implication underlying the binding of cognate and noncognate peptide ligands to their domain receptors. It is expected that the newly proposed method could be further used to perform genome-wide inference of domain-peptide binding at three-dimensional structure level.