Back to the program.
Advances in stochastic local search for protein structure prediction
Holger H. Hoos
Computer Science Department
University of British Columbia


In this talk, I will report on two studies recently completed in my group in which we introduced and studied new stochastic local search methods for solving simplified but prominent formulations of the well-known problem of energy-based prediction of tertiary protein structure. In the first part, I will motivate and describe a new replica-exchange Monte Carlo algorithm for the widely studied 2D- and 3D-hydrophobic-polar models. In the second part, I will present a new adaptive search scheme for the conformations of beta-sheet homopolymers on the face-centred cubic lattice. This latter scheme is quite general and can in principle be applied to many other combinatorial optimisation problems. The goal of the talk is to introduce the audience to computational protein structure prediction as well as to the algorithmic approaches that enabled us to improve the state of the art of solving the specific problems we studied.


protein structure prediction, stochastic local search