Generative Modeling of Full-Atom Protein Conformations using Latent Diffusion on Graph Embeddings

2025/06/20 by Aditya Sengar, Sengar, Aditya, Ali Hariri +7 · 2 voices · 3 citations
Biochemistry, Genetics and Molecular Biology · Computer Science · #Computational Drug Discovery Methods #Protein Structure and Dynamics #Receptor Mechanisms and Signaling #cs.LG #q-bio.BM

paper · pdf · doi:10.48550/arxiv.2506.17064

openalex publication_date 2025/06/20 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Generating diverse, all-atom conformational ensembles of dynamic proteins such as G-protein-coupled receptors (GPCRs) is critical for understanding their function, yet most generative models simplify atomic detail or ignore conformational diversity altogether. We present latent diffusion for full protein generation (LD-FPG), a framework that constructs complete all-atom protein structures, including every side-chain heavy atom, directly from molecular dynamics (MD) trajectories. LD-FPG employs a Chebyshev graph neural network (ChebNet) to obtain low-dimensional latent embeddings of protein conformations, which are processed using three pooling strategies: blind, sequential and residue-based. A diffusion model trained on these latent representations generates new samples that a decoder, optionally regularized by dihedral-angle losses, maps back to Cartesian coordinates. Using D2R-MD, a 2-microsecond MD trajectory (12 000 frames) of the human dopamine D2 receptor in a membrane environment, the sequential and residue-based pooling strategy reproduces the reference ensemble with high structural fidelity (all-atom lDDT of approximately 0.7; C-alpha-lDDT of approximately 0.8) and recovers backbone and side-chain dihedral-angle distributions with a Jensen-Shannon divergence of less than 0.03 compared to the MD data. LD-FPG thereby offers a practical route to system-specific, all-atom ensemble generation for large proteins, providing a promising tool for structure-based therapeutic design on complex, dynamic targets. The D2R-MD dataset and our implementation are freely available to facilitate further research.

Citations

• All-atom Diffusion Transformers: Unified generative modelling of molecules and materials
• Proteina: Scaling Flow-based Protein Structure Generative Models
• Scalable emulation of protein equilibrium ensembles with generative deep learning
• P2DFlow: A Protein Ensemble Generative Model with SE(3) Flow Matching
• Boltz-1 Democratizing Biomolecular Interaction Modeling
• Generative Modeling of Molecular Dynamics Trajectories
• Sequence-Augmented SE(3)-Flow Matching For Conditional Protein Backbone Generation
• Accurate structure prediction of biomolecular interactions with AlphaFold 3
• Boosting Diffusion Models with Moving Average Sampling in Frequency Domain
• Protein Conformation Generation via Force-Guided SE(3) Diffusion Models
• AlphaFold Meets Flow Matching for Generating Protein Ensembles
• Predicting mutational effects on protein-protein binding via a side-chain diffusion probabilistic model
• De novo design of protein structure and function with RFdiffusion
• Str2Str: A Score-based Framework for Zero-shot Protein Conformation Sampling
• A Latent Diffusion Model for Protein Structure Generation
• Evolutionary-scale prediction of atomic-level protein structure with a language model
• Flow Matching for Generative Modeling
• Protein structure generation via folding diffusion
• Protein structure generation via folding diffusion
• Equiformer: Equivariant Graph Attention Transformer for 3D Atomistic Graphs
• Understanding over-squashing and bottlenecks on graphs via curvature
• Highly accurate protein structure prediction with AlphaFold
• E(n) Equivariant Graph Neural Networks
• Equivariant message passing for the prediction of tensorial properties and molecular spectra
• Denoising Diffusion Probabilistic Models
• SE(3)-Transformers: 3D Roto-Translation Equivariant Attention Networks
• On the Bottleneck of Graph Neural Networks and its Practical Implications
• Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• SciPy 1.0--Fundamental Algorithms for Scientific Computing in Python
• DrugBank 5.0: a major update to the DrugBank database for 2018
• CHARMM-GUIMembrane Buildertoward realistic biological membrane simulations

Cited by

• AI‐Physics‐Experiment Trinity for Integrated Protein Dynamics Modeling
• Beyond Ensembles: Simulating All-Atom Protein Dynamics in a Learned Latent Space
• Spectral Diffusion for Protein Dynamics

Discussions

• 🧪 New preprint: "Generative Modeling of Full-Atom Protein Conformations using Latent Diffusion on Graph Embeddings" We generate complete, all-atom protein conformational ensembles, capturing both bac [bsky, 12 points, 1 comments]
• Aditya Sengar, Ali Hariri, Daniel Probst, Patrick Barth, Pierre Vandergheynst: Generative Modeling of Full-Atom Protein Conformations using Latent Diffusion on Graph Embeddings https://arxiv.org/abs/2 [bsky, 0 points, 0 comments]

Related

• DiffBP: Generative Diffusion of 3D Molecules for Target Protein Binding
• Accelerating Inference in Molecular Diffusion Models with Latent Representations of Protein Structure
• Beyond Ensembles: Simulating All-Atom Protein Dynamics in a Learned Latent Space
• Constraint Decoupled Latent Diffusion for Protein Backmapping
• Diffusion-Driven Generative Framework for Molecular Conformation Prediction