# Equivariant 3D-Conditional Diffusion Models for Molecular Linker Design

**Authors:** Ilia Igashov, Hannes Stärk, Clément Vignac, Victor Garcia Satorras, Pascal Frossard, Max Welling, Michael Bronstein, Bruno Correia  
**arXiv:** [2210.05274](https://arxiv.org/abs/2210.05274)  
**Date:** 2022-10-11  
**Summary page:** https://www.summarizepaper.com/en/paper/2210.05274v1/

## Key points

- Fragment-based drug discovery is widely used in early-stage drug development
- DiffLinker is a proposed E(3)-equivariant 3D-conditional diffusion model for designing linkers in fragment-based drug discovery
- DiffLinker can connect disconnected fragments and design molecules that incorporate all initial fragments
- It can link an arbitrary number of fragments together and determine the number of atoms in the linker and its attachment points
- DiffLinker outperforms other existing methods by generating more diverse and synthetically-accessible molecules on standard datasets
- Experimental tests show that DiffLinker successfully generates valid linkers conditioned on target protein pockets
- DiffLinker presents a promising solution for designing molecular linkers in fragment-based drug discovery, making it valuable for early-stage drug development.

## Summary

Fragment-based drug discovery is a widely used approach in early-stage drug development. To address the challenge of designing linkers that connect disconnected molecular fragments to create chemically-relevant candidate drug molecules, a team of researchers has proposed DiffLinker, an E(3)-equivariant 3D-conditional diffusion model. DiffLinker is capable of placing missing atoms between disconnected fragments and designing a molecule that incorporates all the initial fragments. It can also link an arbitrary number of fragments together and automatically determine the number of atoms in the linker and its attachment points to the input fragments. The researchers demonstrated that DiffLinker outperforms other existing methods on standard datasets by generating more diverse and synthetically-accessible molecules. Additionally, experimental tests on real-world applications showed that it successfully generates valid linkers conditioned on target protein pockets. Overall, DiffLinker presents a promising solution for designing molecular linkers in fragment-based drug discovery due to its ability to connect multiple disconnected fragments and generate diverse molecules. This makes it a valuable tool for early-stage drug development.