Anisotropic DBSCAN for 3D SMLM Data Clustering.

Area of Science:

• Biophysics
• Cell Biology
• Microscopy

Background:

• Single-molecule localization microscopy (SMLM) enables super-resolution imaging, reconstructing 3D cellular structures.
• Axial localization precision in SMLM is often limited by anisotropic point spread functions, potentially distorting cellular structures.
• Current structure identification methods like DBSCAN typically assume isotropic search volumes.

Purpose of the Study:

• To develop and validate an anisotropic DBSCAN algorithm for improved structure identification in SMLM data.
• To address the challenge of anisotropic localization precision inherent in 3D SMLM techniques.
• To enhance the accuracy of cellular structure analysis by accounting for imaging anisotropies.

Main Methods:

• Simulated ground truth datasets were used to compare anisotropic and isotropic DBSCAN.
• Experimental localization precisions were incorporated to optimize search parameters via computational grid search.
• Anisotropic DBSCAN performance was evaluated under varying localization precision conditions.

Main Results:

• Anisotropic DBSCAN demonstrated more reliable identification of anisotropic clusters compared to standard DBSCAN.
• Optimized search parameters were proposed based on experimental localization precisions.
• The algorithm showed enhanced performance and robustness across different localization precision levels.

Conclusions:

• Anisotropic DBSCAN offers a more rigorous approach to identifying cellular clusters, especially with astigmatism-based 3D SMLM.
• The method accounts for the inherent anisotropic localization precision in SMLM, leading to more accurate structural analysis.
• This advancement is expected to improve the identification of cellular structures in super-resolution microscopy.