Full-length de novo protein structure determination from cryo-EM maps using deep learning.

Area of Science:

• Structural Biology
• Computational Biology
• Biophysics

Background:

• Cryo-electron microscopy (cryo-EM) generates numerous 3D maps, but atomic model building is challenging.
• A gap exists between available cryo-EM maps and reconstructed 3D structures.
• Automated, accurate reconstruction of protein structures from cryo-EM maps is urgently needed.

Purpose of the Study:

• To develop a deep learning-based framework for de novo atomic-accuracy protein structure determination from cryo-EM maps.
• To address the limitations of manual and semi-automatic model building from cryo-EM data.
• To improve the efficiency and accuracy of protein structure reconstruction.

Main Methods:

• A deep learning framework, DeepMM, was developed using Densely Connected Convolutional Networks.
• The method predicts main-chain and Cα positions, amino acid, and secondary structure types directly from cryo-EM maps.
• DeepMM employs a semi-automatic de novo structure determination approach.

Main Results:

• DeepMM achieved atomic accuracy in building all-atom models from cryo-EM maps at near-atomic resolution.
• Extensive validation on simulated and experimental datasets (up to 2931 maps) demonstrated superior performance.
• The algorithm showed significant improvements in accuracy and coverage compared to state-of-the-art methods like RosettaES, MAINMAST, and Phenix.

Conclusions:

• DeepMM offers a highly effective and generalizable solution for automated protein structure determination from cryo-EM maps.
• The method significantly advances the field by bridging the gap between cryo-EM maps and atomic models.
• DeepMM demonstrates the power of deep learning in tackling complex structural biology challenges.