CryoTransformer: A Transformer Model for Picking Protein Particles from Cryo-EM Micrographs

CryoTransformer is a powerful and accurate particle-picking framework using Residual Network (ResNet) and Transformer. CryoTransformer was trained using the largest diverse labelled CryoPPP dataset for the first time. It recognizes and extracts abundance amount of true protein particles from the input micrographs while maintaining low false-positive rates. We performed rigorous evaluation, comparing our method with existing AI based methods and showcasing its robustness through multiple labels of evaluation. Our model outperformed the current state-of-the art methods and is poised to greatly facilitate the automation of Cryo-EM particle recognition.

Comparison of CryoTransformer with existing AI-Based Methods (EMPIAR ID: 10532)

https://github.com/jianlin-cheng/CryoTransformer/assets/24986485/857f3bb3-60be-404c-87e9-ba1978d9126a

Overview of CryoTransformer Pipeliine

Installation

Clone project

git clone https://github.com/jianlin-cheng/CryoTransformer.git
cd CryoTransformer/

Download trained models

curl https://calla.rnet.missouri.edu/CryoTransformer/pretrained_model.tar.gz --output pretrained_model.tar.gz
tar -xvf pretrained_model.tar.gz
rm pretrained_model.tar.gz

Download training data (if required)

curl https://calla.rnet.missouri.edu/CryoTransformer/train_val_test_data.tar.gz --output train_val_test_data.tar.gz
tar -xvf train_val_test_data.tar.gz
rm train_val_test_data.tar.gz

Download test data

curl https://calla.rnet.missouri.edu/CryoTransformer/test_data.tar.gz --output test_data.tar.gz
tar -xvf test_data.tar.gz
rm test_data.tar.gz

Create conda environment

conda env create -f environment.yml
conda activate CryoTransformer

Prediction on Test Data

This code generates the predicted proteins encircled in Micrographs along with the .box and .star files.

python predict.py 

Optional Arguments:
    --save_micrographs_with_encircled_proteins', default='Y', type=str, choices=['Y', 'N'], help='Plot predicted proteins on Micrographs')
    --empiar, default='10081', type=str, help='EMPIAR ID for prediction. It accepts Micrographs in both .mrc or .jpg format')

Example usage: 
    python predict.py --empiar '10081'

Training and Reproducing Results (if required)

python train.py

Optional Arguments:
    --epochs, default=300, type=int, help='Number of training epochs')
    --batch_size, default=8, type=int, help='Batch size for training')
    --device, default='cuda:0', type=str, help='CUDA device for GPU acceleration')
    --remarks, default='CryoTransformer_github', type=str, help='Additional remarks')
    --data_path, default='CryoTransformer/train_val_test_data', type=str, help='Path to the training dataset')

Example Usage:
    python train.py --epochs 300 --batch_size 16  

Training Data Statistics

Data statistics for Training, validating, and Testing CryoTransformer (* Theoretical weight)

Independent Test Data Statistics

Data statistics used for independent Testing sourced from EMPIAR repository

Data statistics used for independent Testing sourced from CryoPPP dataset (* Theoretical weight)

Rights and Permissions

Open Access \ This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Cite this work

If you use the code or data associated with this research work or otherwise find this data useful, please cite:

CryoTransformer

@article {Dhakal2024, \ author = {Ashwin Dhakal and Rajan Gyawali and Liguo Wang and Jianlin Cheng}, \ title = {CryoTransformer: A Transformer Model for Picking Protein Particles from Cryo-EM Micrographs}, \ year = {2024}, \ doi = {10.1093/bioinformatics/btae109}, \ publisher = {Oxford University Press}, \ journal = {Bioinformatics} \ URL = { https://doi.org/10.1093/bioinformatics/btae109 } }

CryoPPP Dataset

@article {Dhakal2023, \ author = {Dhakal, Ashwin and Gyawali, Rajan and Wang, Liguo and Cheng, Jianlin}, \ title = {A large expert-curated cryo-EM image dataset for machine learning protein particle picking}, \ year = {2023}, \ volume = {10}, \ issue = {1}, \ doi = {10.1038/s41597-023-02280-2}, \ journal = {Scientific Data}, \ url = { https://doi.org/10.1038/s41597-023-02280-2 } }