Generative Multi-Objective Bayesian Optimization with Scalable Batch Evaluations for Sample-Efficient De Novo Molecular Design

2025-12-19

AI Summary

Multi-objective de novo molecular design faces challenges from the vastness of chemical space and the computational expense of high-fidelity simulations. To address this, we propose a modular “generate-then-optimize” framework: first, deep generative models (VAEs or GANs) rapidly construct a large candidate pool; second, we introduce qPMHI—a novel batch acquisition function—to efficiently select molecular batches with maximal Pareto-front expansion potential in discrete molecular space. qPMHI enables exact, scalable multi-point optimization, decouples generation from optimization, and avoids pitfalls of continuous latent-space modeling, thereby significantly improving both diversity and sample efficiency. Our method integrates multi-objective Bayesian optimization, Monte Carlo probability estimation, and hypervolume-based improvement metrics. On synthetic benchmarks and a real-world application—discovery of quinone-based cathode materials for aqueous flow batteries—it surpasses state-of-the-art methods within few iterations, achieving simultaneous optimization of performance and structural diversity.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Deep Generative Models & Autoencoders
• Constraint Satisfaction and Optimization: Mixed Discrete/Continuous Optimization

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Designing molecules that must satisfy multiple, often conflicting objectives is a central challenge in molecular discovery. The enormous size of chemical space and the cost of high-fidelity simulations have driven the development of machine learning-guided strategies for accelerating design with limited data. Among these, Bayesian optimization (BO) offers a principled framework for sample-efficient search, while generative models provide a mechanism to propose novel, diverse candidates beyond fixed libraries. However, existing methods that couple the two often rely on continuous latent spaces, which introduces both architectural entanglement and scalability challenges. This work introduces an alternative, modular "generate-then-optimize" framework for de novo multi-objective molecular design/discovery. At each iteration, a generative model is used to construct a large, diverse pool of candidate molecules, after which a novel acquisition function, qPMHI (multi-point Probability of Maximum Hypervolume Improvement), is used to optimally select a batch of candidates most likely to induce the largest Pareto front expansion. The key insight is that qPMHI decomposes additively, enabling exact, scalable batch selection via only simple ranking of probabilities that can be easily estimated with Monte Carlo sampling. We benchmark the framework against state-of-the-art latent-space and discrete molecular optimization methods, demonstrating significant improvements across synthetic benchmarks and application-driven tasks. Specifically, in a case study related to sustainable energy storage, we show that our approach quickly uncovers novel, diverse, and high-performing organic (quinone-based) cathode materials for aqueous redox flow battery applications.

Problem

• Optimizes multi-objective molecular design with conflicting goals
• Enables scalable batch selection for efficient candidate evaluation
• Improves discovery of novel materials for sustainable energy applications

Innovation

• Generative model constructs diverse candidate molecule pool
• qPMHI acquisition function selects optimal batch via ranking
• Scalable batch evaluations enable efficient Pareto front expansion