Machine learning for cluster analysis of localization microscopy data.

Area of Science:

• Biophysics
• Computational Biology
• Microscopy

Background:

• Accurate quantification of molecular clustering in single-molecule localization microscopy (SMLM) is crucial for understanding molecular spatial relationships.
• Existing computational methods struggle with large datasets, sample heterogeneity, and subjective parameters.

Purpose of the Study:

• To develop a supervised machine learning approach for fast and accurate cluster analysis in SMLM data.
• To overcome limitations of existing methods in processing large-scale and heterogeneous SMLM datasets.

Main Methods:

• A supervised machine learning model (neural network) was trained on simulated clustered SMLM data.
• The model classifies millions of points from SMLM datasets to identify molecular clusters.
• The approach allows for refinement of cluster area, shape, and point density measurements.

Main Results:

• The developed machine learning approach demonstrates high speed and accuracy in cluster analysis.
• The method successfully processed simulated and experimental SMLM data.
• Potential for incorporating additional classifiers to identify different cluster subtypes.

Conclusions:

• Supervised machine learning offers an efficient and accurate solution for SMLM cluster analysis.
• This method can handle large datasets and complex biological samples.
• The approach provides refined measurements of cluster properties, aiding biological interpretation.