Deep Evolutionary Learning for Molecular Design
This work addresses the problem of designing novel molecules with improved properties for drug discovery and materials science, offering an incremental improvement over existing optimization methods.
This paper introduces a deep evolutionary learning (DEL) process for molecular design, combining a fragment-based deep generative model with multi-objective evolutionary computation. The method enables evolutionary operations in the latent space and fine-tunes the generative model with high-quality samples, leading to improved property distributions and outperforming multi-objective Bayesian optimization algorithms on two public datasets.
In this paper, we propose a deep evolutionary learning (DEL) process that integrates fragment-based deep generative model and multi-objective evolutionary computation for molecular design. Our approach enables (1) evolutionary operations in the latent space of the generative model, rather than the structural space, to generate novel promising molecular structures for the next evolutionary generation, and (2) generative model fine-tuning using newly generated high-quality samples. Thus, DEL implements a data-model co-evolution concept which improves both sample population and generative model learning. Experiments on two public datasets indicate that sample population obtained by DEL exhibits improved property distributions, and dominates samples generated by multi-objective Bayesian optimization algorithms.