Installation#

brisc supports Linux, macOS, and Windows on Python 3.9+. Install it with conda or pip:

conda (recommended)

conda install -c conda-forge brisc

pip

pip install brisc

conda is recommended because it also sets up the fast MKL BLAS and some of the R packages brisc uses (both covered below); with pip you must handle those yourself.

R packages#

brisc bridges to R through ryp, which uses the R arrow package to move data between Python and R, so arrow is needed for any R interop. The other R packages are feature-specific: Seurat for converting Seurat objects, limma for differential expression, and SingleCellExperiment for converting SingleCellExperiment objects.

Installing brisc with conda (conda install -c conda-forge brisc) sets up R, arrow, and Seurat automatically. With pip you set them up yourself: first install R — CRAN has per-platform instructions — then add arrow and Seurat from an R session with install.packages(c("arrow", "Seurat")). The optional limma and SingleCellExperiment packages are installed separately, only if you need them.

If you need differential expression, install limma:

conda (Linux/macOS)

conda install -c bioconda bioconductor-limma

BiocManager (all platforms)

R -e 'if (!require("BiocManager", quietly = TRUE)) install.packages("BiocManager"); BiocManager::install("limma")'

If you need to convert to or from SingleCellExperiment objects, install it:

conda (Linux/macOS)

conda install -c bioconda bioconductor-singlecellexperiment

BiocManager (all platforms)

R -e 'if (!require("BiocManager", quietly = TRUE)) install.packages("BiocManager"); BiocManager::install("SingleCellExperiment")'

BLAS and threading#

A few key steps — nearest-neighbor search, harmonization, and label transfer — rely on BLAS. On machines with x86 processors (most Linux and Windows machines), conda allows SciPy to be installed with MKL BLAS, which is highly optimized. (You can install this manually with conda install "libblas=*=*mkl" scipy, although installing brisc through conda takes care of this for you.)

However, pip’s SciPy comes with OpenBLAS, which is less optimized and only supports up to 64 threads. To check which backend you have:

import brisc
from threadpoolctl import threadpool_info
print(sorted({pool['internal_api'] for pool in threadpool_info()}))

mkl is the fast path; openblas means the 64-thread cap applies.