CryoREAD: de novo structure modeling for nucleic acids in cryo-EM maps using deep learning

Area of Science:

• Structural Biology
• Computational Biology
• Biophysics

Background:

• DNA and RNA structures are crucial for cellular functions.
• Cryo-electron microscopy (cryo-EM) is increasingly used to determine these structures.
• Nucleic acid structure modeling from cryo-EM maps, especially at lower resolutions, remains a significant challenge.

Purpose of the Study:

• To develop a fully automated de novo DNA/RNA atomic structure modeling method.
• To address the scarcity of computational tools for nucleic acid structure modeling.
• To improve the accuracy of atomic models derived from cryo-EM data.

Main Methods:

• Development of CryoREAD, a deep learning-based method for DNA/RNA structure modeling.
• Utilizing deep learning to identify atomic positions (phosphate, sugar, base) within cryo-EM maps.
• Automated tracing and 3D structure generation from identified atomic positions.

Main Results:

• CryoREAD demonstrated substantially higher accuracy in building atomic models compared to existing methods.
• The method was effective on cryo-EM maps with resolutions ranging from 2.0 to 5.0 Å.
• Successful application to cryo-EM maps of SARS-CoV-2 biomolecular complexes.

Conclusions:

• CryoREAD offers a powerful, automated solution for de novo DNA/RNA atomic structure modeling from cryo-EM data.
• The method significantly enhances the ability to model nucleic acid structures, particularly at resolutions coarser than atomic level.
• CryoREAD has potential applications in studying viral complexes and other nucleic acid-related biological systems.