Molecular Generator for Multi-objective Optimization Based on the Pareto Algorithm
Dawei Feng · MEDS Clinical Medicine · 2023
Designing molecules with certain physicochemical features can promote the discovery and optimization of lead compounds. However, most molecular generation models optimize only one physicochemical property, which is not sufficient to determine the availability of a drug. This is because the availability of a drug substance molecule depends on the combined effect of many physicochemical properties. In this study, the pareto method was conducted to optimize the compounds for multi-target molecular characteristics in close approximation to those of the reference compound. In addition, we similarly used the random SMILES method involving amplification and diversification of molecules. Finally, we further examined the generation ability of the model and also analyzed the probability distribution of the physicochemical attributes and molecular structure of the created compounds. We expect that the model could develop additional molecules for exploring a bigger chemical space for medicinal chemists.