E3-CryoFold: One-shot Prediction For Cryo-EM Structure Determination

E3-CryoFold is a deep learning framework for automating the determination of three-dimensional atomic structures from high-resolution cryo-electron microscopy (Cryo-EM) density maps. It addresses the limitations of existing AI-based methods by providing an end-to-end solution that integrates training and inference into a single streamlined pipeline. E3-CryoFold combines 3D and sequence Transformers for feature extraction and employs an equivariant graph neural network to build accurate atomic structures from density maps.

Table of Contents

• Background
• Features
• Installation
• Quick Start
• Usage
  • Command-Line Arguments
  • Running the Example
  • Using Custom Data
• Tutorial
• References
• Contact
• License

Background

Cryo-electron microscopy (Cryo-EM) has revolutionized structural biology by enabling the visualization of complex biological molecules at near-atomic resolution. The technique generates high-resolution density maps that offer insights into the molecular structures of proteins, viruses, and other biomolecular assemblies. However, interpreting these density maps to derive accurate atomic models remains a challenging and labor-intensive task, often requiring expert knowledge and manual interventions.

Existing AI-based methods for automating Cryo-EM structure determination face several limitations:

1. Multi-stage processing: Current approaches often involve separate stages for feature extraction, sequence alignment, and structure prediction, leading to inefficiencies and discontinuities.
2. Alignment bias: Techniques such as Hidden Markov Models (HMMs) or Traveling Salesman Problem (TSP) solvers introduce bias when aligning predicted atomic coordinates with the protein sequence.
3. Poor generalization: Due to the limited size of available datasets, many methods struggle to generalize well to complex or previously unseen test cases.

E3-CryoFold addresses these challenges by providing a fully integrated, end-to-end solution that performs one-shot inference with minimal manual intervention, enabling faster and more accurate structure determination.

Features

• 🚀 End-to-End Training and Inference: Simplifies the process by seamlessly integrating training and inference into a single, unified framework, eliminating the need for multi-stage processing.
• ⚡ Fast and Accurate: Achieves a 400% improvement in TM-score over Cryo2Struct while reducing inference time by a factor of 1,000.

For more details on the performance and benchmarking, please refer to our paper.

Installation

To get started with E3-CryoFold, follow these steps:

1. Clone the repository:

   git clone https://github.com/A4Bio/E3-CryoFold.git
   cd E3-CryoFold

2. Create and activate the conda environment:

   conda env create -f environment.yml
   conda activate cryofold

3. Download the Pretrained Model:

   We provide a pretrained model for E3-CryoFold. Download it here and place it in the pretrained_models directory.

4. Download the Experimental dataset: The training set can be downloaded in https://doi.org/10.7910/DVN/FCDG0W, and the standard test dataset can be downloaded in https://doi.org/10.7910/DVN/2GSSC9.

Quick Start

To quickly try out E3-CryoFold using an example dataset, run the following command:

bash run_example.sh

This script runs the inference.py script with sample data provided in the examples folder. It uses a sample density map and a ground truth PDB file for evaluation.

We also provide an example tutorial in quick_start.ipynb.

Usage

Command-line Arguments

The inference.py script supports several command-line arguments:

Argument	Description	Default
--density_map_path	Path to the input density map directory (required).	None
--pdb_path	Path to the ground truth PDB file (optional).	None
--model_path	Path to the pretrained model checkpoint.	pretrained_model/checkpoint.pt
--output_dir	Directory to save the output PDB file.	results
--device	Device to run the model on (cpu or cuda).	cuda
--verbose	Enable verbose output for debugging.	Disabled

Running the Example

You can run the example directly from the command line:

python inference.py --density_map_path examples/density_map --pdb_path examples/5uz7.pdb

Using Custom Data

To use E3-CryoFold with your own data, you need to provide a Cryo-EM density map and, optionally, a PDB file for evaluating the predicted structure. For example:

python inference.py --density_map_path /path/to/your/density_map --pdb_path /path/to/your/ground_truth.pdb --output_dir /path/to/save/results --device cuda

Tutorial

1. Preprocessing Density Maps:

To normalize your density maps, run:

# Normalize you density maps
$ bash run_data_preparation.bash examples/

After preprocessing, the directory structure should look like:

The organization of the downloaded models should look like:

E3-CryoFold
├── examples
│   ├── density_map
│   │   ├── map.map
│   │   ├── seq_chain_info.json
│   │   └── normed_map.mrc
|   |── pretrained_model
│   │   ├── checkpoint.pt

2. Running Inference:

python inference.py --density_map_path examples/density_map --pdb_path examples/5uz7.pdb 

After inference, the output will be saved in the specified output directory:

E3-CryoFold
├── results
│   └── output.pdb

References

For a complete description of the method, see:

TBD

Contact

Please submit any bug reports, feature requests, or general usage feedback as a github issue or discussion.

• Jue Wang ([email protected])
• Cheng Tan ([email protected])
• Zhangyang Gao ([email protected])

License

This project is licensed under the MIT License. See the LICENSE file for details.