Spindle: Scalable Indexing of Spatial-Omics Data for Fast Querying

Spindle enables atlas-scale spatial transcriptomics searches for gene co-expression patterns by indexing spatial gene covariance matrices, compressing them 113–197× and answering queries in 2.8–28 ms per CPU core.

• Why it matters: Atlas-scale spatial omics makes covariance-based gene relationship searches slow due to high dimensionality, extreme sparsity, and variability, creating a bottleneck for repeatedly accessing second-order statistics.
• What they did: Spindle builds a platform-agnostic index by exploiting approximate block-diagonal structure in neighborhood gene–gene covariance, selectively storing informative components for compact, searchable representations.
• The result: Across 87,000–817,000 cells, the index is 113–197× smaller than full covariance matrices; queries run 5.1–16× faster than scanning, recover a median 81% of exact near neighbors, and in breast cancer match cell-type composition (median difference 0.15 vs 0.54 random).

AI summary of the abstract · may contain errors · not medical advice · read the paper.