CR-I-TASSER: assemble protein structures from cryo-EM density maps using deep convolutional neural networks

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Cryo-electron microscopy (cryo-EM) is crucial for protein structure determination.
• Accurate atomic modeling from cryo-EM density maps remains a significant challenge.

Purpose of the Study:

• To introduce CR-I-TASSER, a novel hybrid method for automated cryo-EM structure determination.
• To enhance the accuracy and robustness of atomic model generation from cryo-EM data.

Main Methods:

• CR-I-TASSER integrates deep neural-network learning with I-TASSER assembly simulations.
• The method utilizes deep learning-based Cα position prediction for improved template quality.
• Fragment assembly simulations are optimized based on predicted Cα positions.

Main Results:

• CR-I-TASSER achieved a correct fold (TM-score >0.5) for 643 out of 778 proteins.
• Performance was 64% higher than other de novo and refinement methods on high-resolution data.
• Deep learning-based Cα prediction significantly boosted final model accuracy.

Conclusions:

• CR-I-TASSER offers a highly accurate and robust method for determining protein structures from cryo-EM data.
• The hybrid approach addresses key challenges in atomic model generation for diverse protein targets and resolutions.
• This work presents a new avenue for advancing cryo-EM structure determination.