Prediction and Classification of Dna Binding Proteins Into Four Major Classes Based on Simple Sequence Derived Features Using Ann

Abstract

The problem of predicting the different classes of DNA binding protein from the protein sequence information is still an open problem in bioinformatics. We implemented a two-layered artificial neural network (ANN) of predicting the DNA binding proteins and their classification into four major classes from their amino-acid sequences. Using 61 sequence derived features we are able to achieve 72.99% correct prediction of proteins into DNA binding/non-DNA binding (in the dataset of 1000 proteins). For the complete set of 61 parameters using 5-fold cross-validated classification, ANN model revealed a superior model (accuracy = 72.99 ± 6.86%, Qpred = 73.952 ± 13.12%, sensitivity = 81.53 ± 6.73% and specificity = 72.54 ± 6.39%). The classification accuracy for predicted DNA binding protein into four sub-classes was 70.73% (on average) using five fold cross validation, indicating that multi-class ANN classification system (61-11-4) may have certain level of unique prediction capability.