Reprogramming Pretrained Target-Specific Diffusion Models for Dual-Target Drug Design

TL;DR

This work forms dual-target drug design as a generative task and curates a novel dataset of potential target pairs based on synergistic drug combinations to transfer the knowledge gained in single-target pretraining to dual-target scenarios in a zero-shot manner.

Abstract

Dual-target therapeutic strategies have become a compelling approach and attracted significant attention due to various benefits, such as their potential in overcoming drug resistance in cancer therapy. Considering the tremendous success that deep generative models have achieved in structure-based drug design in recent years, we formulate dual-target drug design as a generative task and curate a novel dataset of potential target pairs based on synergistic drug combinations. We propose to design dual-target drugs with diffusion models that are trained on single-target protein-ligand complex pairs. Specifically, we align two pockets in 3D space with protein-ligand binding priors and build two complex graphs with shared ligand nodes for SE(3)-equivariant composed message passing, based on which we derive a composed drift in both 3D and categorical probability space in the generative process. Our algorithm can well transfer the knowledge gained in single-target pretraining to dual-target scenarios in a zero-shot manner. We also repurpose linker design methods as strong baselines for this task. Extensive experiments demonstrate the effectiveness of our method compared with various baselines.

Figures (4)

• Figure 1: Overview of our method for dual-target drug design. (a) Illustration of CompDiff and DualDiff. We first align two pockets in 3D space with protein-ligand binding prior and build two complex graph with shared ligand nodes. We then compose the SE(3)-equivariant message to derive the drift on output level (CompDiff) or at each layer of the equivariant neural network (DualDiff). Based on the composed drift, we can generate dual-target ligand molecules by compositional reverse sampling. (b) Illustration of repurposing linker design methods for dual-target drug design. We first identity binding-related fragments from the reference molecules for each of the dual targets and then apply linker design methods to link the fragments and derive a complete molecule that can bind to the dual targets separately.
• Figure 2: RMSD between docked poses towards dual targets of different methods.
• Figure 3: Reference molecules and examples of ligand molecules by different methods generated for the dual targets (UniProt ID: P18507 (top) and Q9UBS5 (bottom)).
• Figure 4: Visualization of more reference molecules and examples designed by TargetDiff, LinkerNet and DualDiff.