brisc: a multithreaded CPU toolkit for single-cell analysis
Preprint
Abstract
Single-cell RNA sequencing datasets have grown from tens of thousands of cells to tens of millions. At this scale, standard toolkits like Scanpy and Seurat need hours per analysis, hindering iterative exploration. We present brisc, the first end-to-end multithreaded CPU toolkit for single-cell analysis, spanning dimensionality reduction, integration, label transfer, clustering, embedding, differential expression and plotting. Unlike GPU-based counterparts, brisc strictly guarantees reproducibility: identical results between runs, regardless of thread count. Algorithmic innovations include a blocked inverted-file-index-based nearest-neighbors algorithm that makes 81–98% fewer errors than Scanpy and Seurat. On a 192-CPU node, brisc ran a standard workflow on 10 million cells in 3 minutes, 116× faster than Scanpy at under half its peak memory usage and 236× faster than Seurat with BPCells. On 96 CPUs, it matched the speed of rapids-singlecell on four H100 GPUs. brisc enables near-real-time, reproducible analysis of tens of millions of cells without specialized hardware.Keywords
This publication has 27 references indexed in Scilit:
- Subnanosecond flash memory enabled by 2D-enhanced hot-carrier injectionNature, 2025
- Integrated multimodal cell atlas of Alzheimer’s diseaseNature Neuroscience, 2024
- Characterizing the impacts of dataset imbalance on single-cell data integrationNature Biotechnology, 2024
- Dictionary learning for integrative, multimodal and scalable single-cell analysisNature Biotechnology, 2023
- Comparison of transformations for single-cell RNA-seq dataNature Methods, 2023
- Scarf enables a highly memory-efficient analysis of large-scale single-cell genomics dataNature Communications, 2022
- How many markers are needed to robustly determine a cell's type?iScience, 2021
- Fast, sensitive and accurate integration of single-cell data with HarmonyNature Methods, 2019
- From Louvain to Leiden: guaranteeing well-connected communitiesScientific Reports, 2019
- The Hungarian method for the assignment problemNaval Research Logistics Quarterly, 1955