Effect of Physiochemical Properties on Classification Algorithms for Kinases Family

Abstract

Kinase phosphorylates specific substrates by transferring phosphate from ATP. These are important targets for the treatment of various neurological disorders, drug addiction and cancer. To organize the kinases diversity and to compare distantly related sequences it is important to classify kinases with high precision. In this study we made an attempt to classify kinases using four different classification algorithms with three different physiochemical features. Our results suggest that Random Forest gives an average precision of 0.99 for classification of kinases; and when amphiphilic pseudo amino acid composition was used as feature, the precision of the classifier was much higher than compared to amino acid composition and dipeptide composition. Hence, Random forest with amphiphilic pseudo amino acid composition is the best combination to achieve classification of kinases with high precision. Further the same can be extended for subfamilies, which can give more insight into the predominant features specific to kinases subfamilies.