OptiSim: An extended dissimilarity selection method for finding diverse representative subsets

Compound selection methods currently available to chemists are based on maximum or minimum dissimilarity selection or on hierarchical clustering. Optimizable K-Dissimilarity Selection (OptiSim) is a novel and efficient stochastic selection algorithm which includes maximum and minimum dissimilarity-based selection as special cases. By adjusting the subsample size parameter K, it is possible to adjust the balance between representativeness and diversity in the compounds selected. The OptiSim algorithm is described, along with some analytical tools for comparing it to other selection methods. Such comparisons indicate that OptiSim can mimic the representativeness of selections based on hierarchical clustering and, at least in some cases, improve upon them.