MLIPDiff: Guiding Denoising Diffusion Models for Structure-Based Drug Design via Machine-Learned Interatomic Potentials

Yiyan Liao; Amin Tavakoli; Yu Zhang; Anima Anandkumar

cs.LGOct 23, 2026

MLIPDiff: Guiding Denoising Diffusion Models for Structure-Based Drug Design via Machine-Learned Interatomic Potentials

Authors: Yiyan Liao, Amin Tavakoli, Yu Zhang, Anima Anandkumar

Organizations: Caltech

Abstract

Diffusion models have become a dominant approach to structure-based drug design. However, because they are trained primarily to fit reference structures, they often place ligands in physically implausible poses that clash with the protein. We introduce MLIPDiff, an inference-time guidance method that can be applied to pretrained diffusion models for ligand generation. The method uses physics-informed guidance from UMA (Universal Models for Atoms), a machine-learned interatomic potential (MLIP). The guidance combines four weighted energy terms (protein-ligand interaction, ligand strain, steric clash, and pocket anchoring) and uses the original model checkpoint without retraining. On CrossDocked2020, MLIPDiff completely eliminates protein-ligand collisions while preserving the ligand's local geometry, improving bond-length distributions and physical plausibility relative to the base model. Compared to models such as NucleusDiff that are specifically designed to reduce atomic clashes, MLIPDiff achieves clash-free structures without compromising local geometry. It maintains the base model's PoseBusters pass rate and attains the best median Vina Score, Vina Min, and Vina Dock, surpassing both the base model and NucleusDiff, with a median of 59.6% of ligands per pocket binding more strongly than the reference. MLIPDiff thus addresses a limitation of prior methods by generating physically plausible structures with favorable predicted binding affinity using a single frozen model and no additional training.

Figures & tables

Figure 1: Overview of MLIPDiff. A sampling trajectory of MLIPDiff (light blue), with guidance turned on when $t$ falls below a threshold. At each step $\phi_{\theta}$ predicts the one-step denoised ligand $[\hat{x}_{0},\hat{v}_{0}]$ (sea green), on which four guidance terms yield a gradient that shifts the mean of the reverse-process posterior. TargetDiff (light salmon), the same backbone without guidance, ends in atomic collisions with the pocket (red circles). NucleusDiff (light green) produces an implausible conformation (red circle).

Bond	liGAN	AR	Pocket2Mol	TargetDiff	NucleusDiff	MLIPDiff (Ours)
C $-$ C	0.601	0.609	0.496	0.295	0.500	0.280
C $=$ C	0.665	0.620	0.561	0.190	0.426	0.186
C $-$ N	0.634	0.474	0.416	0.238	0.395	0.230
C $=$ N	0.749	0.635	0.629	0.163	0.460	0.170
C $-$ O	0.656	0.492	0.454	0.297	0.458	0.290
C $=$ O	0.661	0.558	0.516	0.397	0.417	0.383

Table 1: Jensen-Shannon divergence between the bond-length distributions of reference and generated molecules (

\downarrow

better). “

-

”, “

=

”, and “:” represent single, double, and aromatic bonds, respectively.

Figure 2: Median RMSD of rigid fragments before and after force-field optimization, by fragment size (number of heavy atoms).

Figure 3: Protein-ligand collision counts across sampling steps, at the atom-pair, atom, and molecule level (log scale). The blue dashed line marks the guidance onset (step 600,

t\leq t_{\text{thr}}=400

Panel of Table 2 — Table 2: Binding affinity and molecular properties on CrossDocked2020 (100 pockets, 100 ligands each). Vina energies in kcal/mol ( $\downarrow$ better); best in bold, second best underlined. For liGAN and GraphBP ( $*$ ), AutoDock Vina could not parse some generated atom types and thus we do not provide their Vina Score and Vina Min results.

Figure 4: Visualization of the pockets and sampled ligands on the CrossDocked2020 test set. Vina scores are provided to indicate binding affinities.

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

$t$	$1-\bar{\alpha}_{t}$	Force (with H)	Force (without H)	$n$
999	0.632	106.0	101.2	0 68
900	0.552	92.3	104.3	0 74
800	0.455	65.4	98.0	0 78
700	0.340	53.4	73.5	0 83
600	0.217	33.4	33.3	0 99
500	0.110	11.9	0 8.8	0 99

Appendix

Table 3: UMA heavy-atom RMS force (eV/Å) on

\hat{x}_{0}

, median over 100 randomly drawn training-set ligands at each noise level

t

n

is the number of those ligands that OpenBabel reconstructs into a valid molecule.

Term	Weight	Fraction
interaction $w_{\text{int}}$	$0.10$	$20\%$
strain $w_{\text{strain}}$	$0.016$	$20\%$
clash $w_{\text{clash}}$	$1.0$	$40\%$
anchor $w_{\text{anchor}}$	$0.40$	$20\%$

Appendix

Table 4: Guidance weights used in all experiments, each calibrated to a fixed fraction of the per-atom diffusion step

\lVert\mu_{\theta}-x_{t}\rVert

t=100

Figure 5: Per-atom guidance displacement across sampling. Each panel plots the applied per-atom move against the diffusion step size (Diff Move); panels (a) to (d) are the four energy terms and (e) their vector sum.

Structure-Based Drug Design

Structure-based drug design uses protein structures and binding-site geometry to find, rank, and create molecules that bind therapeutic targets.

See all 45 papers

FuseDiff: Symmetry-Preserving Joint Diffusion for Dual-Target Structure-Based Drug Design
Mar 5, 2026Jianliang Wu, Anjie Qiao, Zhen Wang +2Structure-Based Drug Design Drug Discovery
School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou, China · School of Computer Science and Engineering, Guangdong Province Key Laboratory of Computational Science, Sun Yat-sen University, Guangzhou, China · Gaoling School of Artificial Intelligence, Renmin University of China, Beijing, China +1
Ligand-Conditioned Discrete Diffusion for Protein Sequence-Structure Co-Design
May 15, 2026Chen Wei, Fanding Xu, Minghao Sun +5Protein Language Models Diffusion Language Model Decoding
Xi’an University of Posts & Telecommunications · National University of Singapore · Xi’an Jiaotong University +1
Fine-tuning Pocket-Aware Diffusion Models via Denoising Policy Optimization
May 17, 2026Yuan Xue, Daniel Kudenko, Megha KhoslaMolecular Optimization Diffusion Model Fine-Tuning
L3S Research Center, Hanover, Germany · Delft University of Technology, Delft, Netherlands

Mar 5, 2026J/K move · Enter open · S save

FuseDiff: Symmetry-Preserving Joint Diffusion for Dual-Target Structure-Based Drug Design

Jianliang Wu, Anjie Qiao, Zhen Wang, Zhewei Wei+1

School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou, China · School of Computer Science and Engineering, Guangdong Province Key Laboratory of Computational Science, Sun Yat-sen University, Guangzhou, China · Gaoling School of Artificial Intelligence, Renmin University of China, Beijing, China+1

Structure-Based Drug Design Drug Discovery 3D Molecular Generation Molecular Graph Generation Conditional Diffusion Models

Open paper AskPDF arXiv

Dual-target structure-based drug design aims to generate a single ligand together with two pocket-specific binding poses, each compatible with a corresponding target pocket, enabling polypharmacological therapies with improved efficacy and reduced resistance. Existing approaches typically rely on staged pipelines, which either decouple the two poses via conditional-independence assumptions or enforce overly rigid correlations, and therefore fail to jointly generate two target-specific binding modes. To address this, we propose FuseDiff, an end-to-end diffusion model that jointly generates a ligand molecular graph and two pocket-specific binding poses conditioned on both pockets. FuseDiff features a message-passing backbone with Dual-target Local Context Fusion (DLCF), which fuses each ligand atom's local context from both pockets to enable expressive joint modeling while preserving the desired symmetries. Together with explicit bond generation, FuseDiff enforces topological consistency across the two poses under a shared graph while allowing target-specific geometric adaptation in each pocket. To support principled training and evaluation, we derive a dual-target training set and use an independent held-out test set for evaluation. Experiments on the benchmark and a real-world dual-target system show that FuseDiff achieves state-of-the-art docking performance and enables the first systematic assessment of dual-target pose quality prior to docking-based pose search.

Abstract

Structure-Based Drug Design

Explore similar work

FuseDiff: Symmetry-Preserving Joint Diffusion for Dual-Target Structure-Based Drug Design

Ligand-Conditioned Discrete Diffusion for Protein Sequence-Structure Co-Design

Fine-tuning Pocket-Aware Diffusion Models via Denoising Policy Optimization