CryoVirusDB: An Annotated Dataset for AI-Based Virus Particle Identification in Cryo-EM Micrographs.

Area of Science:

• Structural biology
• Virology
• Computational biology

Background:

• Single-particle cryo-electron microscopy (cryo-EM) enables atomic resolution 3D virus structure determination.
• Virus structures are vital for understanding biological functions and developing antiviral therapies.
• AI development for virus particle identification in cryo-EM is limited by a lack of labeled datasets.

Purpose of the Study:

• Introduce CryoVirusDB, a comprehensive dataset for training AI models.
• Facilitate accurate identification and extraction of virus particles from cryo-EM images.
• Advance the field of automated 3D viral structure determination.

Main Methods:

• Compiled 9941 cryo-EM micrographs from nine datasets.
• Included seven distinct non-enveloped viruses with icosahedral symmetry.
• Annotated 339,398 virus particles with expert-picked coordinates.

Main Results:

• CryoVirusDB provides a large-scale, expertly labeled resource for cryo-EM data.
• The dataset enables training and benchmarking of AI and machine learning algorithms.
• Demonstrates potential for improving automated virus particle picking.

Conclusions:

• CryoVirusDB addresses the critical need for labeled data in AI-driven cryo-EM.
• Facilitates the development of robust AI tools for viral structure analysis.
• Aims to accelerate the process of building atomic 3D structural models for viruses.