Benchmarking sketching methods on spatial transcriptomics data.

High-throughput spatial transcriptomics (ST) now profiles hundreds of thousands of cells or locations per section, creating computational bottlenecks for routine analysis. Sketching, or intelligent sub-sampling, addresses scale by selecting small, representative subsets. While effective for single-cell RNA sequencing data, existing sketching methods, which optimize coverage in expression space but ignore physical location, can introduce spatial bias when applied to ST data. To explore the impact of sketching on ST analysis, we systematically benchmarked uniform sampling, leverage-score sampling, Geosketch (minimax/Hausdorff), and scSampler (maximin) across multiple real ST datasets (mouse ovary, MERFISH brain, human breast cancer, lung) and simulations, using three input representations: Principle Component Analysis (PCA) embeddings, spatial coordinates, and spatially smoothed embeddings. We show that expression-only designs capture global transcriptomic heterogeneity but distort tissue architecture by over-sampling high-variability regions and under-sampling homogeneous areas. Coordinate-only sampling restores tissue coverage but misses transcriptional extremes. A simple spatially aware extension, computing leverage scores from a randomized singular value decomposition (SVD) basis smoothed by a spatial weights matrix, strikes a favorable balance, recovering rare cell states while maintaining uniform tissue coverage and avoiding edge effects. Across robust Hausdorff distances, clustering stability (Adjusted Rand Index), PCA loading drift, and local cell-type mean squared error (MSE), spatially smoothed leverage scores match or outperform alternatives. These results motivate joint spatial-transcriptomic sketching objectives to enable fast, unbiased analyses of increasingly large ST datasets.