Prediction Research Of G-protein Coupled Receptor And Its Function

G-protein coupled receptors (GPCRs) constitute the largest family of transmembrane proteins. Identifying novel GPCRs and elucidating their functions from protein primary sequences can not only facilitate to understand signal transduction process and elucidate disease mechanism, but also provide valuable clues for drug R&D, which has become an important research issue both in theoretical and practical aspects. However, elucidating the functions of these proteins by traditional experiment methods is not only time-consuming, but also difficult to keep up with increasing sequences. There is a wider gap between sequence data and functional knowledge of GPCRs proteins. Therefore, using computational methods to predict GPCRs and their functions is one of the effective ways to solve this problem. The main contributions are summarized as follows:Firstly, the features of protein sequence including dipeptide composition, amino acid composition, length of sequence, and physiochemical properties of amino acid such as hydrophobicity, polarity, charge of amino acids, have been analyzed and extracted effectively. Based on these features, classifiers are constructed for family classification and functional prediction of GPCRs with the method of support vector machine (SVM). The cross-validation results demonstrate that GPCRs could be correctly identified with an accuracy of 99% at family classification and 97% GPCRs can be identified correctly, respectively. The prediction performances are better than previous methods. Then, large-scale analysis is performed for GPCRs discovery in the human genome based on GPCR-PRED, a predicting system for GPCRs classification, prediction and coupling specificity prediction based on this strategy, some meaningful results are further obtained. GPCR-PRED is available in the website (http://moe.hgrp.cn) and can be accessed freely.Secondly, the features information of physiochemical properties of amino characters are extracted, and then convert them into vectors with fixed length. Classifiers are designed for predicting cytokine family classification and prediction for cytokine, which represents one of important ligand families of GPCRs. A predicting tool, CYTOKINE-PRED, is developed based on this method. The results of cross-validation demonstrate that cytokine could be correctly identified with an accuracy of 90% at family classification and 95% cytokine can be identified correctly, respectively.Furthermore, aiming at the problems of coupling regions selection and feature recognition of current methods, features based on physiochemical properties and composing of amino acid residue in the whole GPCRs sequences are extracted, and then convert them into vectors with fixed length. Classifiers are designed for predicting G protein coupling specificity. The results of cross-validation demonstrate that the system of GPCR-PRED could achieve an accuracy of 90% at coupling prediction. The prediction performances are better than previous methods.Finally, the web-based chemokine information database and GPCRs drug target database are developed by integrating and extracting reliable data sources associated with chemokine and drug target. The interaction information between chemokines and their receptors provided by the chemokine database can facilitate to identify important metabolic pathways. The drug target database can reveal insights into the identification and validation of GPCRs targets. They are available in the website (http://moe.hgrp.cn) and can be accessed freely.