Cryo2StructData: A Large Labeled Cryo-EM Density Map Dataset for AI-based Modeling of Protein Structures

Abstract

The advent of single-particle cryo-electron microscopy (cryo-EM) has brought forth a new era of structural biology, enabling the routine determination of large biological molecules and their complexes at atomic resolution. The high-resolution structures of biological macromolecules and their complexes significantly expedite biomedical research and drug discovery. However, automatically and accurately building atomic models from high-resolution cryo-EM density maps is still time-consuming and challenging when template-based models are unavailable. Artificial intelligence (AI) methods such as deep learning trained on limited amount of labeled cryo-EM density maps generate inaccurate atomic models. To address this issue, we created a dataset called Cryo2StructData consisting of 7,600 preprocessed cryo-EM density maps whose voxels are labelled according to their corresponding known atomic structures for training and testing AI methods to build atomic models from cryo-EM density maps. Cryo2StructData is larger than existing, publicly available datasets for training AI methods to build atomic protein structures from cryo-EM density maps. We trained and tested deep learning models on Cryo2StructData to validate its quality showing that it is ready for being used to train and test AI methods for building atomic models.

De novo atomic protein structure modeling for cryoEM density maps using 3D transformer and HMM

“…The models were trained as a sequence-to-sequence predictor, utilizing a Transformer-Encoder 17 to capture long-range voxel-voxel dependencies and a skip-connected decoder to combine the extracted features at different encoder layers to classify each voxel. The models were trained using the large Cryo2StructData dataset 21 . Cryo2StructData is a comprehensive labeled dataset of cryo-EM density maps curated specifically for deep learning-based atomic structure modeling in cryo-EM density maps.…”

“…The dataset used to train and validate Cryo2Struct (Cryo2StructData) is available on the Harvard Dataverse 24 , and the description of the data preparation and labeling process can be found in 21 . The detailed information about the test datasets including the EMD IDs of the density maps and the evaluation scores are provided in two Excel files (Standard_test_data.xlsx for the standard test dataset and Cryo2Struct_test_data.xlsx for the new test dataset) available at 10.7910/DVN/GQCTTD, and the true structures and the structural models built by Cryo2Struct and Phenix for the test density maps are also available at the same website.…”