Cheminformatics

Sesame: Structure-Aware Molecular Generation via Spatial Density-Map Conditioning

Konstantin Yatsenko, Arvind Thiagarajan

Featured July 5, 2026

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

Sesame is a smart molecular designer that can either invent new drug-like molecules from scratch or grow existing partial molecules by understanding 3D shapes, using a special density map and diffusion process.

In depth

Sesame introduces a novel diffusion-based molecular generation model that leverages a spatial pairformer module to condition on 3D structural information, such as protein pockets or partial ligand fragments. This conditioning is achieved by representing these structures as continuous spatial density maps, enabling a unified approach for both *de novo* molecule generation and fragment-conditioned lead optimization.

Key Takeaways

1
The paper proposes a novel density map conditioning architecture that adaptively samples from 3D density maps to guide molecular generation, making it structure-aware.
2
A unified conditioning mechanism supports both *de novo* generation and fragment-conditioned lead optimization by treating protein pockets and partial ligands as the same continuous density field.
3
The model employs a hybrid discrete-continuous diffusion process to jointly denoise atom types, bond types, and atomic coordinates within a single framework.

Conceptual Flow

HIGH LEVEL

Methodology: How Sesame Generates Molecules

Sesame takes a jumbled molecule and a 3D map of its environment, then slowly cleans up the molecule step-by-step until it's a complete, good-fitting drug.

Jumbled Molecule

3D Environment Map

Clean Up Step-by-Step

Complete Drug Molecule

Results: Impact on Drug Discovery

By combining new molecule design with growing existing ones, Sesame helps create many valid, drug-like compounds that fit well into target areas.

Drug Design Problem

Sesame's Unified Solution

Valid Drug Molecules

Preserves Key Parts

This breakdown was generated by SciGrove. Get the same analysis — intuition, storyboard, peer review, a runnable prototype and a glossary — on any paper you upload or paste a DOI for.