Deciphering the Association of Single Amino Acid Variations with Dermatological Diseases Applying Machine Learning Techniques

Abstract

SAVDerma predicts the association of a Single Amino Acid Variation (SAVs) with dermatological diseases. SAVs are basically non-synonymous SNPs which are sometimes associated with various diseases. A single amino acid mutation can affect various physico-chemical properties of a protein which then effect the normal functionality of a particular protein and leads to disease. Studies have shown that among various mendelian diseases, around 60% of them are due to these types of SAVs. The scope of this study limits with dermatological disorders and the process can be applied to all other types of diseases. We have curated SAV’s physico-chemical and sequence-based features which are associated with dermatological disorders and kept them as positive set and vice versa as negative set. This data sets were feed to machine learning classifiers for developing a model which can classify SAV’s association with dermatological disorders. Our classifier obtained an accuracy of 87.29%. SAVDerma is a web application where user can find all the curated and machine generated data about SAVs and their predicted association with skin diseases. It has a user-friendly interface where information regarding particular SAV can be retrieved using three different parameters i.e. Gene symbol, Swissprot id and rs id and are provided on the search page of the website (http://savderma.info/). SAVDerma can be used as a single point information retrieval system by clinicians and thus increasing the current knowledge on skin disorders.