SpatialLeiden: spatially aware Leiden clustering.

Area of Science:

• Computational Biology and Bioinformatics
• Single-cell and Spatial Omics Analysis

Background:

• Clustering algorithms identify inherent structures in data, crucial for annotating cell types in single-cell omics.
• Leiden clustering is widely adopted in single-cell analysis but often considered non-spatial.
• Spatial omics data retains information about molecular phenotypes and their spatial organization within tissues.

Purpose of the Study:

• To demonstrate that Leiden clustering can be adapted into a spatially aware method for spatial omics data.
• To evaluate the performance of this adapted Leiden clustering against existing state-of-the-art spatial clustering algorithms.

Main Methods:

• Integration of spatial information into the Leiden clustering algorithm at multiple stages.
• Computational implementation of a spatially aware Leiden clustering approach.
• Benchmarking against established spatial clustering techniques using relevant metrics.

Main Results:

• The modified Leiden clustering demonstrates high computational performance.
• The method effectively incorporates spatial context, becoming 'spatially aware'.
• Performance is competitive with, and comparable to, current leading spatial clustering algorithms.

Conclusions:

• Leiden clustering can be successfully enhanced with spatial awareness for spatial omics applications.
• This adaptation offers a computationally efficient and effective alternative for spatial data analysis.
• The findings expand the utility of a popular clustering algorithm to the growing field of spatial omics.