brisc: a multithreaded CPU toolkit for single-cell analysis

brisc, a multithreaded CPU toolkit for single-cell RNA-seq, analyzes 10 million cells in 3 minutes with strict run-to-run reproducibility across thread counts.

• Why it matters: Standard single-cell toolkits like Scanpy and Seurat take hours at tens-of-millions scale, limiting iterative exploration and making reproducible, fast workflows difficult.
• What they did: brisc provides an end-to-end multithreaded CPU pipeline covering dimensionality reduction, integration, label transfer, clustering, embedding, differential expression, and plotting, using a blocked inverted-file-index nearest-neighbors method.
• The result: On a 192-CPU node, it is 116x faster than Scanpy and 236x faster than Seurat with BPCells while using under half peak memory, and its nearest-neighbors approach yields 81-98% fewer errors.

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