Docking-based Multi-objective Molecular optimization Pipeline using Structure-constrained Genetic Algorithm

Yurim Lee, Kyudam Choi, Cheongwon Kim

Abstract: Lead optimization is a multi-objective optimization problem that satisfies various pharmacological properties by considering the binding affinity with a disease protein target. Maintaining the substructure to have selectivity for the disease protein target while satisfying the pharmacological properties is essential. optimizing and ranking multiple candidate molecules is problematic because it involves conflicting and incomparable goals that must be maximized or minimized simultaneously. To solve this problem, we introduce a method of structure-constrained molecular optimization using a genetic algorithm with molecular docking. The proposed algorithm performs multi-objective optimization for multi-properties while satisfying the scaffold constraint of the molecular structure through two-phase optimization. The experiments of BACE1 inhibitor optimization indicate that the proposed model can effectively discover molecules that conserve scaffold substructure and satisfy pharmacological properties. Our model is available at https://github.con/devYurim/DMMOPGA.