Abstract:Understanding the protein structures is invaluable in various biomedical applications, such as vaccine development. Protein structure model building from experimental electron density maps is a time-consuming and labor-intensive task. To address the challenge, machine learning approaches have been proposed to automate this process. Currently, the majority of the experimental maps in the database lack atomic resolution features, making it challenging for machine learning-based methods to precisely determine protein structures from cryogenic electron microscopy density maps. On the other hand, protein structure prediction methods, such as AlphaFold2, leverage evolutionary information from protein sequences and have recently achieved groundbreaking accuracy. However, these methods often require manual refinement, which is labor intensive and time consuming. In this study, we present DeepTracer-Refine, an automated method that refines AlphaFold predicted structures by aligning them to DeepTracers modeled structure. Our method was evaluated on 39 multi-domain proteins and we improved the average residue coverage from 78.2 to 90.0% and average local Distance Difference Test score from 0.67 to 0.71. We also compared DeepTracer-Refine with Phenixs AlphaFold refinement and demonstrated that our method not only performs better when the initial AlphaFold model is less precise but also surpasses Phenix in run-time performance.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to improve the accuracy of cryo - electron microscopy (cryo - EM) structure prediction by combining the deep learning methods DeepTracer and AlphaFold2. Specifically, the paper proposes a new method named DeepTracer - Refine, aiming to address the following two main challenges: 1. **Limitations of the density - map - based modeling method (Map - to - Model)**: - **Resolution limitations**: Most cryo - EM density maps obtained from experiments lack atomic - level resolution features, making it difficult for machine - learning methods to accurately determine protein structures. - **Residue identification and connection errors**: Due to the limited resolution of cryo - EM density maps, machine - learning algorithms have difficulty in accurately classifying side chains, resulting in inaccurate prediction of amino acid types. In addition, due to the noise and quality differences in experimental maps, methods such as DeepTracer may have incorrect connections when connecting residues. 2. **Limitations of the sequence - based modeling method (Sequence - to - Model)**: - **Folding accuracy in cross - domain regions**: Although AlphaFold2 has made breakthrough progress in predicting protein structures, it still has accuracy problems when folding the regions between protein domains. These regions are usually flexible and disordered, so AlphaFold2's predictions in these regions are not precise enough. To overcome these challenges, the paper proposes the DeepTracer - Refine method to improve the prediction results of AlphaFold2 through the following steps: - **Utilizing the high - sequence - coverage of AlphaFold2**: AlphaFold2 can provide complete sequence coverage, but the predicted backbone geometry may be inaccurate. - **Combining DeepTracer's high - resolution density - map analysis**: DeepTracer can identify residues from high - resolution cryo - EM density maps and generate preliminary structure predictions. - **Automatic segmentation and alignment**: DeepTracer - Refine detects low - confidence regions in AlphaFold2 predictions, segments the protein structure into compact domains, and aligns these domains with DeepTracer's prediction results to improve the overall structure's accuracy. Through this method, DeepTracer - Refine significantly improves the residue coverage and local distance difference test (lDDT) score in the structure prediction of multi - domain proteins, thus solving the above - mentioned challenges.

Enhancing cryo-EM structure prediction with DeepTracer and AlphaFold2 integration

Beyond Current Boundaries: Integrating Deep Learning and AlphaFold for Enhanced Protein Structure Prediction from Low-Resolution Cryo-EM Maps

Integrating AlphaFold and deep learning for atomistic interpretation of cryo-EM maps

Protein complex structure modeling by cross-modal alignment between cryo-EM maps and protein sequences

DeepTracer: Fast Cryo-EM Protein Structure Modeling and Special Studies on CoV-related Complexes

Protein model refinement for cryo-EM maps using AlphaFold2 and the DAQ score

Improved automated model building for cryo-EM maps using CryFold

DeepMainmast: integrated protocol of protein structure modeling for cryo-EM with deep learning and structure prediction

DiffModeler: Large Macromolecular Structure Modeling in Low-Resolution Cryo-EM Maps Using Diffusion Model

A Tool for Segmentation of Secondary Structures in 3D Cryo-EM Density Map Components Using Deep Convolutional Neural Networks

DiffModeler: large macromolecular structure modeling for cryo-EM maps using a diffusion model

Fast and Automated Protein-DNA/RNA Macromolecular Complex Modeling from Cryo-EM Maps

Accurate flexible refinement for atomic-level protein structure using cryo-EM density maps and deep learning

CR-I-TASSER: assemble protein structures from cryo-EM density maps using deep convolutional neural networks

AlphaFold2 transmembrane protein structure prediction shines

A New Protocol for Atomic-Level Protein Structure Modeling and Refinement Using Low-to-Medium Resolution Cryo-EM Density Maps

Predictive modeling and cryo-EM: A synergistic approach to modeling macromolecular structure

Full-lengthde Novoprotein Structure Determination from Cryo-Em Maps Using Deep Learning

High-accuracy protein structures by combining machine-learning with physics-based refinement

Model Building of Protein Complexes from Intermediate-Resolution Cryo-Em Maps with Deep Learning-Guided Automatic Assembly

Accounting Conformational Dynamics into Structural Modeling Reflected by Cryo-EM with Deep Learning