A Winning Score-Based Evolutionary Process For Multi-And Many-Objective Peptide Optimization

Target identification as part of drug design is a long process with high laboratory evaluation costs since optimal candidate leads have to be identified in an iterative process including the determination of diverse physiochemical properties, which have to be optimized simultaneously. MOEAs have become an established optimization method in in silico-aided drug design processes. Since target identification becomes more complex, the dimension of molecular optimization problems increases. Less work has been done so far to evolve an evolutionary process efficiently solving both, multi- and many-objective molecular optimization problems while considering application-specific conditions of molecule optimization. This work presents the enhancement of a MOEA especially evolved for molecular optimization. The proposed algorithm is applicable to multi- and many-objective molecular optimization problems identifying a selected number of qualified candidate peptides within a very low number of iterations. It has a simple framework structure and optionally uses two types of winning-score ranking method as survival selection. Default parameters are provided in the components to enable a non-expert use. This algorithm is benchmarked to the recently proposed and promising AnD (ANgle-based selection and shift-based Density estimation strategy) on molecular optimization problems up to 6 objectives. Furthermore, the selection principles are exemplarily compared and discussed.