Machine Learning Method to Differentiate Ataxias

Abstract

Spinocerebellar ataxias or SCAs, are a group of more than 37 genetically and clinically heterogeneous known neurodegenerative diseases. This work analyzes the level of genetic similarity between several ataxias, we identified proteins that are associated with more than one ataxia. A decision tree was trained to identify ataxias by identifying whether a new entry disease not yet identified and not classified can be grouped as an ataxia. Altogether 12 proteins from different ataxias were verified, all 12 proteins were analyzed in 500 different trees to better evaluate the method used. Of the 12 proteins tested, the method was correct for 10 different proteins or 83% of correct results. This identifier and the results obtained in the experiments allow a greater characterization of the diseases addressed, it also allows applications such as the reuse of treatments for similar diseases.