A comparison of contributions to disease phenotype between damaging and benign non-synonymous SNPs

Abstract

Non-synonymous SNPs (nsSNPs), also known as Single Amino acid Polymorphisms (SAPs), are likely to affect the function of the proteins accounting for susceptibility to complex disease for their altering the encoded amino acid sequence. Recent advances in genetic studies found that the non-synonymous variations locating in disordered regions are functionally important. We therefore considered predicting deleterious SAPs based on both protein interaction network and disordered protein property. We used one of functional prediction algorithms of nsSNPs, PolyPhen-2, to distinguish SAPs as damaging or benign. Four classifiers: naive Bayes, k-Nearest Neighbor (kNN), Support Vector Machine (SVM) and Random Forests (RF) were used to classify SAPs. As a result, the prediction accuracies of four classifiers are all over 70%, and the three features (degree, clustering coefficient and disorder score) were found to be potential predictor variables to classify nsSNPs.