CAGenMol Condition-Aware Molecule Diffusion Overview

Paper context

Paper title: CAGenMol: Condition-Aware Diffusion Language Model for Goal-Directed Molecular Generation Abstract: Goal-directed molecular generation requires satisfying heterogeneous constraints such as protein--ligand compatibility and multi-objective drug-like properties, yet existing methods often optimize these constraints in isolation, failing to reconcile conflicting objectives (e.g., affinity vs. safety), and struggle to navigate the non-differentiable chemical space without compromising structural validity. To address these challenges, we propose CAGenMol, a condition-aware discrete diffusion framework over molecular sequences that formulates molecular design as conditional denoising guided by heterogeneous structural and property signals. By coupling discrete diffusion with reinforcement learning, the model aligns the generation trajectory with non-differentiable objectives while preserving chemical validity and diversity. The non-autoregressive nature of diffusion language model further enables iterative refinement of molecular fragments at inference time. Experiments on structure-conditioned, property-conditioned, and dual-conditioned benchmarks demonstrate consistent improvements over state-of-the-art methods in binding affinity, drug-likeness, and success rate, highlighting the effectiveness of our framework. Passages referencing this figure: l . CAGenMol: Condition-Aware Diffusion Language Model for Goal-Directed Molecular Generation Yanting LI 1, † † thanks: These authors contributed equally to this work. , Zhuoyang JIANG 1 1 1 footnotemark: 1 , Enyan DAI 1 , Lei WANG 2 , Wen-Cai Ye 2 , Li LIU 1 , 1 The Hong Kong University of Science and Technology (Guangzhou), 2 Jinan University, Guangzhou Correspondence: avrillliu@hkust-gz.edu.cn Figure 1: Overview of CAGenMol. UCA encodes either protein-pocket structure or target properties, which guides a condition-aware masked diffusion model on SAFE sequences. The model is further trained with Step-PPO, and perform inference with EFO refinement. 1 Introduction The discovery of novel small-molecule therapeutics is a cornerstone of modern medicine DiMasi et al. ( 2016 ); Hughes et al. ( AFE’s fragment-based representation imposes strong chemical priors, preventing local invalidity even under aggressive optimization. We further initialize our framework with the pre-trained GenMol Lee et al. ( 2025 ) backbone to inherit learned chemical distributions, allowing the model to focus exclusively on condition alignment rather than learning basic validity. 4 Methodology As illustrated in Figure 1 , we present a unified framework CAGenMol for goal-directed molecular generation that synergizes condition-aware discrete diffusion with reinforcement learning. The framework is designed to explicitly align molecular generation with complex biochemical objectives beyond pure data distribution modeling. 4.1 Model Architecture. The core design philosophy of CAGenMol is to bridge the modalit