CryoFSL: An Annotation-Efficient, Few-Shot Learning Framework for Robust Protein Particle Picking in Cryo-EM...

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Accurate protein particle identification in cryo-electron microscopy (cryo-EM) is essential for high-resolution structure determination.
• Current methods often require extensive annotated datasets and struggle with low signal-to-noise ratio (SNR) conditions, limiting generalization to new protein targets.

Purpose of the Study:

• To develop a novel few-shot learning framework for robust and annotation-efficient particle picking in cryo-EM.
• To significantly reduce the annotation burden while maintaining or improving performance compared to existing methods.

Main Methods:

• Developed CryoFSL, a few-shot learning framework utilizing Segment Anything Model 2 (SAM2) with lightweight adapters.
• Implemented a hierarchical adapter design for dynamic feature modulation to handle low-SNR and heterogeneous cryo-EM data.
• Evaluated the framework using minimal labeled micrographs (as few as five) across diverse protein targets.

Main Results:

• CryoFSL achieved superior recall, precision, and 3D reconstruction resolution compared to traditional and state-of-the-art deep learning methods in a few-shot setting.
• The framework demonstrated robustness and stability across heterogeneous micrographs, consistently identifying high-quality particles with fewer false positives.
• Achieved competitive density map reconstruction resolution using a fraction of the particles compared to other methods.

Conclusions:

• CryoFSL offers a scalable, generalizable, and annotation-efficient solution for particle picking in cryo-EM.
• This approach significantly reduces the reliance on large annotated datasets, making cryo-EM analysis more accessible and efficient.
• The framework redefines efficiency and quality standards in cryo-EM data processing.