Protein model assessment via machine learning techniques

Abstract

We attempt to solve the problem of protein model assessment using machine learning techniques and information from sequence and structure of the protein. The goal is to generate a machine that understands structures from PDB and given a new model, predicts whether or not it belongs to the class of PDB structures. We show two such machines (SVM and FDT); results appear promising for further analysis. To reduce computational overhead, multiprocessor environment and basic feature selection method is used. The prediction accuracy using improved FDT is above 80% and results are better when compared with other machine learning techniques.