Incorporating multiple biology based knowledge to amplify the prophecy of enzyme sub-functional classes

Abstract

Based on current in silico methods, enzyme sub-functional classes is distinguished from sequence level information, local order or sequence length and order knowledge. To date, no work has been done to predict the enzyme subclasses efficiently corresponding to the ENZYME database. In order to precisely predict the sub-functional classes of enzyme, we propose a derivative feature vector labelled as APH which unifies amino acid composition, dipeptide composition, hydrophobicity and hydrophilicity. Support Vector Machine is used for prediction and the performance is evaluated using accuracy obtained over 99% and Matthew's Correlation Coefficient (MCC) over 0.99 with the aid of biological validation from in vivo studies.