Multi-objective latent space optimization of generative molecular design models

• DOI: 10.1016/j.patter.2024.101042
• Publication Date: 2024-08-12T14:34:58Z
• Authors: A N M Nafiz Abeer; Nathan M. Urban; M. Ryan Weil; Francis J. Alexander; Byung-Jun Yoon
• Keywords: Generative model; Retraining

Molecular design based on generative models, such as variational autoencoders (VAEs), has become increasingly popular in recent years due to its efficiency for exploring high-dimensional molecular space identify molecules with desired properties. While the efficacy of initial model strongly depends training data, sampling suggesting novel enhanced properties can be further via latent optimization (LSO). In this paper, we propose a multi-objective LSO method that significantly enhance performance (GMD). The proposed adopts an iterative weighted retraining approach, where respective weights data are determined by their Pareto efficiency. We demonstrate our GMD improve jointly optimizing multiple