sketchR provides a simple interface to the geosketch and
scSampler
python packages, which implement subsampling algorithms described in
Hie et al (2019)
and Song et al (2022),
respectively. The implementation makes use of the
basilisk
package for interaction between R and python.
You can install sketchR from Bioconductor (release 3.19 onwards) using:
if (!require("BiocManager", quietly = TRUE))
install.packages("BiocManager")
BiocManager::install("sketchR")library(sketchR)
## Create an example data matrix. Rows represent "samples" (the unit of
## downsampling), columns represent features (e.g., principal components).
mat <- matrix(rnorm(5000), nrow = 500)
## Run geosketch. The output is a vector of indices, which you can use
## to subset the rows of the input matrix.
idx <- geosketch(mat, N = 100)
## Run scSampler. As for geosketch, the output is a vector of indices.
idx2 <- scsampler(mat, N = 100)