MolRGen: A Training and Evaluation Setting for De Novo Molecular Generation with Reasonning Models
This addresses a gap in drug discovery for researchers by providing a supervised training setting for de novo molecular generation, though it is incremental as it builds on existing reasoning LLM approaches.
The paper tackles the lack of training setups for de novo molecular generation with reasoning-based LLMs by introducing MolRGen, a benchmark and dataset, and demonstrates its use by training a 24B LLM with reinforcement learning to generate novel molecules.
Recent advances in reasoning-based large language models (LLMs) have demonstrated substantial improvements in complex problem-solving tasks. Motivated by these advances, several works have explored the application of reasoning LLMs to drug discovery and molecular design. However, most existing approaches either focus on evaluation or rely on training setups that require ground-truth labels, such as molecule pairs with known property modifications. Such supervision is unavailable in \textit{de novo} molecular generation, where the objective is to generate novel molecules that optimize a desirability score without prior knowledge of high-scoring candidates. To bridge this gap, we introduce MolRGen, a large-scale benchmark and dataset for training and evaluating reasoning-based LLMs on \textit{de novo} molecular generation. Our contributions are threefold. First, we propose a setting to evaluate and train models for \textit{de novo} molecular generation and property prediction. Second, we introduce a novel diversity-aware top-$k$ score that captures both the quality and diversity of generated molecules. Third, we show our setting can be used to train LLMs for molecular generation, training a 24B LLM with reinforcement learning, and we provide a detailed analysis of its performance and limitations.