Chapter 3 - Machine Learning for Protein Structure and Function Prediction

Abstract

Abstract Solving biological problems in silico has posed many daunting challenges ranging from simulating the protein dynamics to function assignment for new sequences. One approach to such problems, machine learning, is to learn the solution from a set of examples. Machine learning has found wide application in many problem domains such as medical diagnosis, market analysis, traffic analysis among others. The key idea in machine learning is to direct the computer to learn how to solve a problem rather than explicitly give the solution to the computer. This chapter reviews a number of popular machine learning formulations and gives an example for each formulation using the modeling of protein–DNA interactions. Then we present recent applications of machine learning to a number of biological problems related to protein structure and function prediction. These include protein-RNA interactions, protein-membrane interactions, protein-protein interactions, protein-peptide interactions, single amino acid polymorphisms, subcellular localization, and protein structure predictions.