A Metropolis Monte Carlo algorithm for merging single-particle diffraction intensities.

Area of Science:

• X-ray science
• Computational imaging
• Biophysics

Background:

• Single-particle imaging (SPI) using X-ray free-electron lasers (XFELs) reconstructs nanoscale structures from diffraction patterns.
• Accurate reconstruction relies on merging numerous weak diffraction patterns, often with missing particle orientation data.
• The expand-maximize-compress (EMC) algorithm is a key method for this data merging process.

Purpose of the Study:

• To enhance the computational efficiency of the EMC algorithm for SPI.
• To address limitations of exhaustive parameter sampling in standard EMC implementations.
• To explore alternative sampling strategies for handling complex reconstruction problems with multiple missing parameters.

Main Methods:

• Modification of the standard EMC algorithm to incorporate Metropolis Monte Carlo (MMC) sampling.
• Comparison of the MMC-based EMC variant with the traditional grid-sampling EMC algorithm.
• Utilizing simulated diffraction data for performance evaluation.

Main Results:

• The MMC-based EMC variant demonstrates potential computational advantages over grid-based EMC.
• This modified algorithm is particularly beneficial for reconstruction problems with more than three missing parameters.
• Simulated data analysis validates the feasibility and efficiency of the proposed method.

Conclusions:

• Metropolis Monte Carlo sampling offers a computationally tractable alternative to grid sampling within the EMC framework for XFEL SPI.
• This algorithmic advancement can facilitate more complex single-particle imaging reconstructions.
• The developed method shows promise for improving the scalability and applicability of SPI techniques.