SynLaD: Latent Diffusion for Generating Synthesizable Molecules Conditioned on 3D Pharmacophore Profiles

2026-07-01

AI Summary

This work addresses the longstanding challenge in molecular generation of simultaneously optimizing three-dimensional (3D) pharmacophore-constrained structures for potency and synthetic feasibility. To this end, the authors propose SynLaD, a novel framework that unifies reaction-constrained retrosynthetic pathway generation and 3D pharmacophore-guided geometric modeling within a single latent diffusion model. SynLaD employs an encoder–dual-decoder architecture—comprising a geometric decoder and an autoregressive synthesis decoder—coupled with a pharmacophore-conditioned diffusion Transformer to jointly refine molecular structures and synthetic routes in latent space. Experimental results demonstrate that SynLaD significantly outperforms existing baselines across multiple bioactive ligand generation tasks, achieving high synthetic accessibility while enhancing both molecular diversity and hit rates.

Abstract

We present SynLaD, a latent diffusion framework for small-molecule generation that unifies ligand-based drug design objectives (what to make) with synthetic accessibility (how to make it). Current models typically optimize one objective at the expense of the other, creating a bottleneck for discovering high-scoring and synthesizable molecules. SynLaD combines reaction-constrained generation with pharmacophore-conditioned 3D design by learning a latent space that decodes to both 3D structures and synthesis pathways. An encoder maps molecules to a latent representation used by two decoder heads: (i) a geometric head that reconstructs atom types and coordinates and (ii) an autoregressive synthesis head that outputs synthetic routes in a serialized, reaction-based notation. A diffusion transformer generates novel latents in the learned space, conditioned on pharmacophore profiles. Across analogue generation tasks for bioactive ligands, SynLaD outperforms existing baselines in synthesizable and diverse hit generation, demonstrating that a single model can produce shape-aligned molecules with feasible synthesis plans.

Problem

• molecule generation
• synthetic accessibility
• pharmacophore
• drug design
• latent diffusion

Innovation

• latent diffusion
• pharmacophore conditioning
• synthetic accessibility
• 3D molecular generation
• reaction-based synthesis