ProFusion: 3D Reconstruction of Protein Complex Structures from Multi-view AFM Images

2025-09-17

AI Summary

To address the challenges of high experimental costs, low computational prediction accuracy, and missing pose annotations in multi-view atomic force microscopy (AFM) images for 3D structural determination of large protein complexes, this paper proposes an end-to-end multi-view 3D reconstruction method that bypasses explicit pose estimation. Our approach features two key innovations: (1) a physics-informed virtual AFM simulation framework to generate large-scale, physically realistic synthetic training data; and (2) a hybrid architecture integrating conditional diffusion models with instance-specific neural radiance fields (NeRFs), enabling direct high-fidelity 3D reconstruction from sparse, noisy AFM images. Evaluated on multiple real protein complex datasets, our method achieves an average Chamfer distance at the AFM imaging resolution limit (~1 nm), significantly outperforming state-of-the-art alternatives. The framework demonstrates high reconstruction accuracy, strong generalizability across diverse complexes, and practical feasibility for experimental structural biology.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Affective Computing

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Web data generation and simulation
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

AI-based in silico methods have improved protein structure prediction but often struggle with large protein complexes (PCs) involving multiple interacting proteins due to missing 3D spatial cues. Experimental techniques like Cryo-EM are accurate but costly and time-consuming. We present ProFusion, a hybrid framework that integrates a deep learning model with Atomic Force Microscopy (AFM), which provides high-resolution height maps from random orientations, naturally yielding multi-view data for 3D reconstruction. However, generating a large-scale AFM imaging data set sufficient to train deep learning models is impractical. Therefore, we developed a virtual AFM framework that simulates the imaging process and generated a dataset of ~542,000 proteins with multi-view synthetic AFM images. We train a conditional diffusion model to synthesize novel views from unposed inputs and an instance-specific Neural Radiance Field (NeRF) model to reconstruct 3D structures. Our reconstructed 3D protein structures achieve an average Chamfer Distance within the AFM imaging resolution, reflecting high structural fidelity. Our method is extensively validated on experimental AFM images of various PCs, demonstrating strong potential for accurate, cost-effective protein complex structure prediction and rapid iterative validation using AFM experiments.

Problem

• Reconstructing 3D protein complex structures from multi-view AFM images
• Overcoming limitations of expensive experimental techniques like Cryo-EM
• Addressing the impracticality of large-scale experimental AFM training data

Innovation

• Hybrid deep learning with AFM imaging
• Virtual AFM framework simulating imaging process
• Conditional diffusion and NeRF for reconstruction