We are continuing our work on porting techniques from evolutionary computation to computational structural biology and protein modeling for the purpose of designing powerful conformational search algorithms. In this direction, we have pursued population-based Evolutionary Search algorithms (EAs) that additionally project sampled conformations to nearby local minima. We have shown that these resulting hybrid (memetic) algorithms are indeed more powerful than basic evolutionary search algorithms (including Basin Hopping) and indeed comparable to state-of-the-art decoy sampling algorithms based on Metropolis Monte Carlo optimization (such as, Rosetta). This work has recently appeared in: 1) Sameh Saleh, Brian Olson, and Amarda Shehu. "A Population-based Evolutionary Algorithm for Sampling Minima in the Protein Energy Surface." Comput Struct Biol Workshop (CSBW), pg. 48-55, Philadelphia, PA, 2012. An extended version of this work is in: 2) Sameh Saleh, Brian Olson, and Amarda Shehu. "A population-based evolutionary search approach to the multiple minima problem in de novo protein structure prediction." BMC Structural Biology Journal, 2013.

A recent collaboration with Prof. Kenneth De Jong is focusing on measuring the efficacy of different reproduction techniques in the context of population-based evolutionary search algorithms for decoy sampling. This work has appeared in: Brian Olson, Kenneth A De Jong, and Amarda Shehu. "Off-Lattice Protein Structure Prediction with Homologous Crossover." GECCO, Amsterdam, Netherlands, 2013.

Brian's presentation at GECCO 2013.