Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Pharmacophore-Conditioned Diffusion Model for Ligand-Based De Novo Drug Design

Amira Alakhdar, Barnabas Poczos, Newell Washburn

TL;DR

The paper presents PharmaDiff, a pharmacophore-conditioned diffusion model for 3D molecular generation that does not require protein structures. Built on a MiDi-inspired SE(3)-equivariant framework, PharmaDiff conditions generation on an atom-based 3D pharmacophore graph via inpainting, COM adjustment, and cross-attention, producing molecules that satisfy predefined pharmacophoric constraints. Across ligand-based and structure-based experiments, PharmaDiff achieves superior pharmacophore matching, higher docking scores, and favorable drug-likeness properties, outperforming key baselines such as PGMG, TransPharmer, REINVENT4, and DiffSBDD in relevant metrics. These results demonstrate the potential of pharmacophore-guided diffusion for efficient, target-agnostic de novo drug design with practical implications for accelerating drug discovery workflows.

Abstract

Developing bioactive molecules remains a central, time- and cost-heavy challenge in drug discovery, particularly for novel targets lacking structural or functional data. Pharmacophore modeling presents an alternative for capturing the key features required for molecular bioactivity against a biological target. In this work, we present PharmaDiff, a pharmacophore-conditioned diffusion model for 3D molecular generation. PharmaDiff employs a transformer-based architecture to integrate an atom-based representation of the 3D pharmacophore into the generative process, enabling the precise generation of 3D molecular graphs that align with predefined pharmacophore hypotheses. Through comprehensive testing, PharmaDiff demonstrates superior performance in matching 3D pharmacophore constraints compared to ligand-based drug design methods. Additionally, it achieves higher docking scores across a range of proteins in structure-based drug design, without the need for target protein structures. By integrating pharmacophore modeling with 3D generative techniques, PharmaDiff offers a powerful and flexible framework for rational drug design.

Pharmacophore-Conditioned Diffusion Model for Ligand-Based De Novo Drug Design

TL;DR

The paper presents PharmaDiff, a pharmacophore-conditioned diffusion model for 3D molecular generation that does not require protein structures. Built on a MiDi-inspired SE(3)-equivariant framework, PharmaDiff conditions generation on an atom-based 3D pharmacophore graph via inpainting, COM adjustment, and cross-attention, producing molecules that satisfy predefined pharmacophoric constraints. Across ligand-based and structure-based experiments, PharmaDiff achieves superior pharmacophore matching, higher docking scores, and favorable drug-likeness properties, outperforming key baselines such as PGMG, TransPharmer, REINVENT4, and DiffSBDD in relevant metrics. These results demonstrate the potential of pharmacophore-guided diffusion for efficient, target-agnostic de novo drug design with practical implications for accelerating drug discovery workflows.

Abstract

Developing bioactive molecules remains a central, time- and cost-heavy challenge in drug discovery, particularly for novel targets lacking structural or functional data. Pharmacophore modeling presents an alternative for capturing the key features required for molecular bioactivity against a biological target. In this work, we present PharmaDiff, a pharmacophore-conditioned diffusion model for 3D molecular generation. PharmaDiff employs a transformer-based architecture to integrate an atom-based representation of the 3D pharmacophore into the generative process, enabling the precise generation of 3D molecular graphs that align with predefined pharmacophore hypotheses. Through comprehensive testing, PharmaDiff demonstrates superior performance in matching 3D pharmacophore constraints compared to ligand-based drug design methods. Additionally, it achieves higher docking scores across a range of proteins in structure-based drug design, without the need for target protein structures. By integrating pharmacophore modeling with 3D generative techniques, PharmaDiff offers a powerful and flexible framework for rational drug design.
Paper Structure (30 sections, 7 equations, 3 figures, 3 tables)

This paper contains 30 sections, 7 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Preliminaries
  4. Denoising diffusion models
  5. Gaussian diffusion for continuous data
  6. Discrete diffusion for categorical data
  7. Diffusion for molecules
  8. Molecular representation
  9. Noise model
  10. Methods
  11. Featurization of pharmacophores
  12. Model overview
  13. Pharmacophore feature encoding
  14. Inpainting
  15. COM adjustment
  16. ...and 15 more sections

Figures (3)

  • Figure 1: PharmaDiff's architecture overview simultaneously predicting 2D and 3D coordinates conditioned on a pharmacophore hypothesis. The model uses 12 layers of an E(3) graph Transformer architecture, designed to maintain SE(3) equivariance, it used inpainting and a cross-attention layer between molecular and pharmacophore Node Embeddings to enforce pharmacophore constrains.
  • Figure 2: (A) Decomposition of molecular structures into 3D pharmacophore-associated atoms. (B) Conditioning the PharmaDiff model during the denoising process using the pharmacophore graph $G_p$. The pharmacophoric features shown include hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), hydrophobic (HYD), and aromatic (ARO) groups.
  • Figure 3: Distribution of physicochemical properties for the GEOM-Drugs training set and molecules generated by PharmaDiff. The plot compares 100,000 molecules generated from random pharmacophore hypotheses with 100,000 conformers randomly selected from the GEOM-Drugs training set.