Use batch balanced KNN (BBKNN) in R

BBKNN is a fast and intuitive batch effect removal tool for single-cell data. It is originally used in the scanpy workflow, and now can be used with Seurat seamlessly.

bbknnR has been tested on R versions >= 4.1. Please consult the DESCRIPTION file for more details on required R packages. bbknnR has been tested on Linux platforms

To use the full features of bbknnR, you also need to install the bbknn python package:

pip install bbknn

bbknnR has been released to CRAN:

install.packages("bbknnR")

or can be installed from github:

devtools::install_github("ycli1995/bbknnR")

library(bbknnR)
library(Seurat)
data("panc8_small")
panc8_small <- RunBBKNN(panc8_small, batch_key = "tech")

• Fix trim <- trim %||% 10 * ncol(nn.idx). The correct code should be trim <- trim %||% (10 * ncol(nn.idx)). (#10)

• Add k_build_nndescent = 30 parameter to match the implementation of python bbknn.

• Remove reticulate dependency. Now use kNN algorithms provided by RcppAnnoy and rnndescent
• Add return.umap.model for RunBBKNN.Seurat
• Improvements for testthat

• Compatibility with Seurat v5
• Improvements for documentation and verbose.

• Explicit import of get_dummies.() from tidytable
• Fix a bug when pass only one batch_key to RidgeRegression()

• Import public function similarity_graph() from uwot==0.1.14 in compute_connectivities_umap() to follow the CRAN policy

• Initially released to CRAN

Please cite this implementation R in if you use it:

Yuchen Li (2022). bbknnR: Use batch balanced KNN (BBKNN) in R.
package version 0.1.0 https://github.com/ycli1995/bbknnR

Please also cite the original publication of this algorithm.

Polanski, Krzysztof, et al. "BBKNN: fast batch alignment of single cell transcriptomes." Bioinformatics 36.3 (2020): 964-965.