Pareto Optimal Flexible Alignment of Molecules using a Non-Dominated Sorting Genetic Algorithm

The use of multiobjective function optimization to superimpose a flexible source molecule onto a rigid target molecule is explored. The objective functions are the SEAL similarity score between the source and target molecules, which has to be maximized, and the conformational strain of the source molecule, which has to be minimized. The optimization algorithm used is an elitist non-dominated sorting genetic algorithm. The algorithm is tested with the superpostition of two non-nucleoside HIV-reverse transcriptase inhibitors and the superposition of methotrexate onto dihydrofolate.
Reprints of this paper can be requested at info@molmo.be or from CILS directly.

schematic overview of the non-dominated sorting genetic algorithm
Schematic overview of the non-dominated sorting genetic algorithm



Scatterplot
Scatterplot of the results of 5 runs of the non-dominated sorting GA applied to the superposition of ITU on DATA. The solid line is a quadratic fit through the Pareto front.

Superposition of methotrexate onto dihydrofolate
Superposition of methotrexate (black) onto dihydrofolate (grey). Left the 'X-ray-mode' alignment, right the 'hetero-mode' alignment. The alignments with the highest SEAL score are shown.