Evolutionary search algorithms for protein modeling

Abstract

In the last two decades, great progress has been made in molecular modelling through computational treatments of biological molecules grounded in evolutionary search techniques. Evolutionary search algorithms (EAs) are gaining popularity beyond exploring the relationship between sequence and function in biomolecules. In particular, recent work is showing the promise of EAs in exploring structure spaces of protein chains and protein assemblies to address open-standing problems in computational structural biology, such as de novo structure prediction and protein-protein docking. Exploring effective interleaving of global and local search has led to hybrid EAs that are now competitive with the Monte Carlo-based frameworks that have traditionally dominated de novo structure prediction. Similar advances have been made in protein-protein docking. Deeper understanding of the constraints posed by highly-coupled modular systems like proteins and integration of domain knowledge has resulted in effective reproductive operators. Multi-objective optimization has also shown promise in dealing with the conflicting terms that make up protein energy functions and effectively exploring protein energy surfaces. Combinations of these techniques have recently resulted in powerful stochastic search frameworks that go beyond de novo structure prediction and are capable of yielding comprehensive maps of possible diverse functionally-relevant structures of proteins. The objective of this tutorial is to introduce the EC community to the rapid developments on EA-based frameworks for protein structure modeling through a concise but comprehensive review of developments in this direction over the last decade. The review will be accompanied with specific detailed highlights and interactive software demonstrations of representative methods. The tutorial will expand the view of EA-based frameworks beyond sequence-focused application settings. The tutorial will introduce EC researchers to open problems in computational structural biology and in the process spur the design of novel and powerful evolutionary search techniques.