Bioinformatics Approach to Classification of Four Classes of Organism in Relation to Their Optimal Growth Temperature

Abstract

 —Identifying the temperature class of proteins in prokaryotic organisms is one of the vital problems in enzyme and protein engineering. In this work, an efficient K-NN predictive models have been developed to discriminate hyperthermophilic, thermophilic, psychrophilic, and mesophilic proteins using Amino acid and Pseudo amino acid compositions. The two predictive models were built and tested with a large dataset consisting of 6631 hyperthermophiles, 11,700 thermophiles, 6267 psychrophiles, and 67,037 mesophiles. Implementation and analysis results showed that the proposed K-NN based predictive models were capable of discriminating the four classes efficiently and with high accuracies, whereby the Amino acid composition model achieved 94% accuracy when using 10-fold cross-validation, and 98% when using hold-out test. on the other hand, the Pseud amino acid composition based model achieved an accuracy of 99% using hold-out test.