Monte carlo methods for small molecule high-throughput experimentation

By analogy with Monte Carlo algorithms, we propose new strategies for desig n and redesign of small molecule libraries in high-throughput experimentati on, or combinatorial chemistry. Several Monte Carlo methods are examined, i ncluding Metropolis, three types of biased schemes, and composite moves tha t include swapping or parallel tempering. Among them, the biased Monte Carl o schemes exhibit particularly high efficiency in locating optimal compound s. The Monte Carlo strategies are compared to a genetic algorithm approach. Although the best compounds identified by the genetic algorithm are compar able to those from the better Monte Carlo schemes, the diversity of favorab le compounds identified is reduced by roughly 60%.