PMut2: a web-based tool for predicting pathological mutations on proteins

Abstract

Amino acid substitutions in proteins can result in an altered phenotype which might lead to a disease. PMut2 is a method that can predict whether a mutation has a pathological effect on the protein function. It uses current machine learning algorithms based on protein sequence derived information. The accuracy of PMut2 is as high as 82%, with a Matthews correlation coefficient of 0,62. PMut2 predictions can be obtained through a modern website which also allows to apply the same machine learning methodology that is used to train PMut2 to custom training sets, allowing users to build their own tailor-made predictors.