Computational Prediction of Amyloidogenic Regions in Protiens : A Machine Learning Approach

Amyloidogenic regions in polypeptide chains are associated with a number of pathologies including neurodegenerative diseases. Recent studies have shown that small regions of proteins are responsible for its amyloidogenic behavior. Therefore, identifying these short peptides is critical for understanding diseases associated with protein aggregation. Owing to the limitations of molecular techniques for the identification of fibril forming targets, it became apparent that clever computational techniques might enable their discovery in silico. We propose a machine learning based method to predict the amyloid fibril-forming short stretches of peptides using Support Vector Machine. The features of this method are based on the physicochemical properties of amino acids. Inorder to get an optimal number of properties, a feature selection approach based on Genetic Algorithm is PErformed. The presented algorithm achieved a balanced prediction performance in terms of true positive and false positive rates in predicting a peptide status: amyloidogenic or non-amyloidogenic, which is not reflected in the existing methods.