Similarity Searching in Files of Three-Dimensional Chemical Structures. Alignment of Molecular Electrostatic Potential Fields with a Genetic Algorithm
David J. Wild, Peter Willett · Journal of Chemical Information and Computer Sciences · 1996
This paper discusses techniques for similarity searching in databases of three-dimensional chemical structures represented by their molecular electrostatic potential fields. Field-based similarity searching involves the alignment of molecular fields so as to maximize their overlap, and we have used a genetic algorithm for this purpose, in which a chromosome encodes the rotations and translations that are needed to generate such an alignment and in which the fitness function is the value of the similarity measure that results from that alignment. The comparison of a pair of typical small-molecules is sufficiently rapid to enable the algorithm to be used with files of nontrivial size. Experiments with a Kendall Square Research KSR-1 multiprocessor system demonstrate the suitability of the algorithm for use in a parallel environment.