spARI

spARI extends the adjusted Rand index to incorporate spatial distance information for evaluating spatial transcriptomics clustering accuracy.

Key Features:

• Spatially aware Rand index (spRI): Weights disagreement object pairs by the distance between the two objects, rather than assigning zero weight as in standard Rand index.
• Spatially aware adjusted Rand index (spARI): Adjusts spRI for random chances such that its expectation equals zero under an appropriate null model.
• Distance-adaptive differentiation: Adaptively differentiates disagreement object pairs based on their distinct distances to favor spatial coherence of clustering.

Scientific Applications:

• Spatial transcriptomics clustering evaluation: Benchmarking and comparing the accuracy of spatial transcriptomics clustering approaches where spatial information is biologically meaningful.
• Simulation studies: Validation of clustering performance under controlled conditions with known ground truth.
• Real ST dataset analysis: Evaluation of clustering methods on experimental spatial transcriptomics datasets to assess spatial coherence.

Methodology:

spRI calculates a weighted disagreement score for object pairs based on their spatial distance, with spARI derived by adjusting spRI for chance agreement; statistical properties are established under a defined null model.