Repairpred a prediction server for proteins of human DNA repair system

Abstract

A DNA undergoes several chemical reactions which result in damage to its structure. This damage is identified and rectified by a collection of processes called DNA repair. The prediction of DNA repair in protein sequences can help in studying the process of DNA repair and the effects its absence can have. It is important to identify whether a certain protein is involved in DNA repair mechanism or not if its effect on DNA repair process has to be studied. In present study, Multilayer perceptron (ANN) and Sequential minimization algorithm (SVM) have been used to create models for the prediction of involvement in DNA repair mechanisms among protein sequences. The sequences used for creation of datasets were extracted from various databases exclusively for DNA repair and non DNA repair for training and testing sets. A significant difference between the accuracy of models generated by ANN and SVM was not obtained, which showed their equal capability to make the prediction. The models were validated using 10-fold cross validation. The final results were compiled and presented in the form of a web server “Repairpred” where users can submit their sequences for prediction in the form of a text file or otherwise.