Skip to content

Reinforcement Learning for Aligning Structure-Based Molecular Generation toward Make-on-Demand Chemical Spaces

Oct 2026 · Journal of Chemical Information and Modeling · 0 citations · 49 references
Computational Drug Discovery Methods

TL;DR

Re reinforcement learning is applied to a pocket-conditioned generative molecular model, demonstrating that RL alignment can substantially improve the experimental accessibility of SBDD candidates without sacrificing structural or chemical quality.

Abstract

Generative models for structure-based drug design have strong potential to accelerate early-stage drug discovery by enabling the conditional generation of target-specific molecules. However, a major bottleneck is that generated compounds are frequently difficult or impossible to synthesize, limiting their direct experimental validation and utility. We address this by optimizing for membership in make-on-demand chemical libraries, ensuring that generated molecules can be directly purchased and tested. To this end, we apply reinforcement learning to a pocket-conditioned generative molecular model, steering generated molecules toward two large make-on-demand spaces: Enamine REAL and Freedom 4.0. We compare an offline, preference-based strategy (direct preference optimization, DPO) and an online RL strategy (DiffusionNFT) across single-target and multi-target settings. In single-target optimization for KRAS, the fraction of unique purchasable candidates rises from 10.0% to 25.5% for Enamine REAL and from 4.3% to 12.1% for Freedom 4.0, while drug-likeness and docking-based affinity proxies are preserved. In multi-target optimization, DPO achieves consistent improvements across diverse protein targets, including a 23-fold increase for the challenging CDK4 pocket. These results demonstrate that RL alignment can substantially improve the experimental accessibility of SBDD candidates without sacrificing structural or chemical quality.

Read PDF

Similar papers

#reinforcement learning Review Open access Oct 2026

Reinforcement Learning for Molecule Design: A Taxonomy of Algorithms, Benchmarks, and Road to Synthesizable Leads

Abstract Reinforcement Learning (RL) has been in the forefront in traversing the vast chemical space for novel drug discovery. In this survey, we discuss the landscape of applying RL in two related paradigms: de novo drug design, the process of generating novel drugs from scratch and drug structure optimization, an app...

Kusal Debnath, Taksheel Shetty, Preetam Ghosh · 0 citations
#artificial intelligence Preprint Oct 2026

ReMedI: Inference-Time Refinement and Self-Distillation for Controllable Molecular Optimization

It is argued a pre-trained model may already contain useful molecular edits that zero-shot generation cannot efficiently reach, and proposed ReMedI, which uses the molecular language model as a base proposer for inference-time alignment, is proposed.

Jia-Yi Zhang, Shivanshu Shekhar, Allison McLevige et al. · 0 citations
#artificial intelligence Preprint Oct 2026

PairIT: Autoregressive Transformers for Low-Data Molecule Optimization

PairIT is introduced, a flexible training framework for autoregressive transformers over SMILES sequences that enables seeded generation by learning from pairs of structurally similar molecules, and Prompt tokens extend this framework by specifying the target property, desired direction, and approximate magnitude of ch...

Hannah Lawrence, Bodhi P. Vani, Ji Won Park et al. · 0 citations
Sep 2026

CRAFT: A Chemical Reaction-Aware Transformer for Synthesizable Molecular Optimization via Multimodal Learning.

CRAFT is a reaction-aware molecular optimization framework that integrates Monte Carlo Tree Search (MCTS) with a multimodal strategy network and explicitly encodes two-dimensional molecular topology and atom-level reaction roles, enabling chemically feasible transformations during multi-step optimization.

Run-Fu Yu, Tong-Mao Ma, Zian Song et al. · 0 citations
Open access Sep 2026

Generative Access to Make-on-Demand Chemical Space Enables Ultra-Large Virtual Screening

Real-SWIT, a generative ULVS framework that learns Enamine REAL Space as a synthesizable molecular distribution and couples this generator with a target-specific scoring model to enable docking-guided exploration of REAL Space at the level of complete molecule, establishes generative access to make-on-demand chemical s...

Kai-Yue Zhang, Ying Sun, Xin-Yue Li et al. · 0 citations
#artificial intelligence Preprint Oct 2026

MFG: Multi-turn Feedback Guided Molecular Optimization

Multi-objective molecular optimization is a critical process in drug discovery that aims to modify a given molecule to simultaneously improve multiple properties while maintaining a similar overall structure. This process often requires iterative modifications guided by computational or experimental feedback. Recent la...

Trieu Nguyen, Frazier N. Baker, Ruo-Xi Gao et al. · 0 citations

Related blog posts

MIT News · Artificial Intelligence Aug 27, 2026

Looking beyond natural sequences

A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.