SynLaD: Latent Diffusion for Generating Synthesizable Molecules Conditioned on 3D Pharmacophore Profiles
2026-07-01 • Machine Learning
Machine Learning
AI summaryⓘ
The authors introduce SynLaD, a new method that helps design small drug-like molecules by considering both what the molecules should do and how they can be made in a lab. Unlike older methods that focus on either designing good molecules or making them easily, SynLaD learns to generate 3D molecule shapes and their possible synthesis steps at the same time. It uses a special model that encodes molecules into a hidden format and then decodes this into both molecule structures and step-by-step recipes. Tests show SynLaD can create diverse, useful molecules with realistic ways to make them.
latent diffusionsmall-molecule generationligand-based drug designsynthetic accessibilitypharmacophore3D molecular designautoregressive synthesisdiffusion transformerreaction-constrained generationanalog generation
Authors
Miruna Cretu, John Bradshaw, Patricia Suriana, Saeed Saremi, Omar Mahmood, Kirill Shmilovich, Kangway Chuang, Vishnu Sresht, Colin Grambow
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.