Ensemble of Artificial Bee Colony Optimization and Random Forest Technique for Feature Selection and Classification of Protein Function Family Prediction

Abstract

Protein function prediction is a prevalent technique in bioinformatics and computational biology. Even now, the computation of function prediction is an impudent task to provide efficient and statistically significant accurate results. In this work, the optimization approach and the machine learning method were proposed to predict the function families of a protein using the sequences regardless of its similarity. It is denoted as Prot-RF (ABC) (predicting protein family using random forest with artificial bee colony). The features of the protein sequences are selected using the ABC method, and they are classified using the random forest classifier. The Uniprot and PDB benchmark databases have been utilized to assess the proposed Prot-RF (ABC) method against the other well-known existing methods such as SVM-Prot, K-nearest neighbor, AdaBoost, probabilistic neural network, Naïve Bayes, random forest, and J48. The classification accuracy results of the proposed Prot-RF (ABC) method outperform the other remaining existing methods.