Simulated annealing: an effective stochastic optimization approach to computational library design.

Abstract

We describe here a stochastic optimization protocol for computational library design based on the principle of simulated annealing (SA). We also demonstrate via computer simulation studies that the SA-guided diversity sampling affords higher information content than random sampling in terms of cluster hit rates. Using a tripeptoid library, we show that the SA guided similarity focusing provides important information about reagent selection for combinatorial synthesis. Finally, we report a system that employs the SA protocol for the simultaneous optimization of multiple properties during library design. We propose that the SA technique is an effective optimization method for computational library design.