Supervised Machine Learning with Protein Structural and Network Topological Features Predicts Physical Interactors of the Human Huntingtons Disease Protein

Abstract

Huntington's disease is a genetically inherited disorder, causing progressive degeneration of the brain. The mutant protein in Huntington's disease patients exhibits complex biophysical properties, and affects numerous cellular processes. Since numerous proteins interact with either the normal or, the mutant huntingtin protein, or both, to decipher the features that enable this discrimination is a complex problem. We trained a Gradient Boosting Machine (GBM) on several protein-structural features and graph-topological features of the normal and the diseased proteins. The GBM was able to achieve an AUC up to 0.88 in predicting interacting partners of the mutant Huntington's disease protein in 10-fold cross-validation trials.