Unsupervised classification of single-particle X-ray diffraction snapshots by spectral clustering

Area of Science:

• Structural Biology
• Biophysics
• X-ray Science

Background:

• Single-particle X-ray Free Electron Laser (XFEL) experiments generate vast amounts of data.
• These datasets often contain a mix of blank shots, single particles, and multi-particle events.
• Experimental data can also be contaminated with unintended molecular species.

Purpose of the Study:

• To develop an automated method for classifying XFEL single-particle snapshots.
• To enable efficient data processing without manual intervention or predefined templates.
• To improve the accuracy and speed of analyzing XFEL experimental outputs.

Main Methods:

• An unsupervised machine learning algorithm was developed.
• The method sorts experimental snapshots without using templates or specific noise models.
• No user-directed learning or prior knowledge of particle types is required.

Main Results:

• The unsupervised method successfully sorts complex mixtures of experimental snapshots.
• The automated classification achieved 90% agreement with manual expert classification.
• This demonstrates the effectiveness of the algorithm in handling real-world experimental data.

Conclusions:

• Unsupervised learning provides a powerful tool for analyzing large XFEL datasets.
• The developed method significantly enhances the efficiency of single-particle analysis.
• This approach facilitates faster and more reliable data interpretation in XFEL research.