Spatially aware adjusted Rand index for evaluating spatial transcriptomics clustering.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Spatial transcriptomics (ST) clustering is vital for understanding tissue heterogeneity.
• Accurate ST clustering enhances downstream biological analysis.
• Current benchmarking methods lack spatial awareness.

Purpose of the Study:

• To propose novel metrics, spRI and spARI, for evaluating ST clustering.
• To address the limitations of the adjusted Rand index (ARI) in spatial data.

Main Methods:

• Developed spatially aware Rand index (spRI) incorporating object distances.
• Introduced spatially aware adjusted Rand index (spARI) with adjustments for random chance.
• Evaluated metrics using simulation studies and real ST datasets.

Main Results:

• spRI and spARI effectively incorporate spatial distance information.
• The proposed metrics favor spatial coherence in clustering.
• spARI demonstrates improved utility over ARI in evaluating ST clustering methods.

Conclusions:

• spRI and spARI offer more accurate evaluations of spatial ST clustering.
• These metrics provide a better assessment of spatial coherence than ARI.
• The proposed methods enhance the benchmarking of ST clustering approaches.