HelixMO: Sample-Efficient Molecular Optimization in Scene-Sensitive Latent Space

Abstract

Efficiently discovering molecules that meet various property requirements can significantly benefit the drug discovery industry. Since it is infeasible to search over the entire chemical space, recent works adopt generative models for goal-directed molecular generation. They tend to utilize the iterative processes, optimizing the parameters of the molecular generative models at each iteration to produce promising molecules for further validation. Assessments are exploited to evaluate the generated molecules at each iteration, providing direction for model optimization. However, most previous works require a massive number of expensive and time-consuming assessments, e.g., wet experiments and molecular dynamic simulations, leading to the lack of practicability. To reduce the assessments in the iterative process, we propose a costeffective evolution strategy in latent space, which optimizes the molecular latent representation vectors instead. We adopt a pre-trained molecular generative model to map the latent and observation spaces, taking advantage of the large-scale unlabeled molecules to learn chemical knowledge. To further reduce the number of expensive assessments, we introduce a pre-screener as the proxy to the assessments. We conduct extensive experiments on multiple optimization tasks comparing the proposed framework to several advanced techniques, showing that the proposed framework achieves better performance with fewer assessments.

Interpretable Machine Learning Model for Predicting Interaction Energies between Dimethyl Sulfide and Potential Absorbing Solvents

“…To get a training set that is representative of the chemical space, more molecules were generated from the initial 46 seed molecules using the sequence variational auto-encoder (SeqVAE) method provided by the computational biology platform of PaddleHelix. 31 Those compounds having a considerable molecular weight (M w > 200), high synthesis difficulty (synthesis accessibility score, SA > 4), or low water solubility (lipid−water partition coefficient, logP > 2.5) were excluded from the initial data set. Finally, 342 molecules were added to the training data set.…”