RankMap

RankMap is a fast, robust, and scalable method for reference-based cell type annotation of single-cell and spatial transcriptomics data. It transforms gene expression matrices into ranked representations based on the top-k expressed genes per cell and applies multinomial regression via the glmnet framework to predict cell types. This rank-based strategy improves robustness to batch effects, platform differences, and partial gene coverage, making RankMap particularly suitable for technologies such as Xenium and MERFISH. The package supports flexible preprocessing, fast model training and prediction, and is compatible with Seurat, SingleCellExperiment, and SpatialExperiment objects. RankMap achieves competitive accuracy with significantly lower runtime than existing methods such as SingleR, Azimuth, and RCTD.

Quick start guide can be found here.

Installation

The RankMap package can be installed from GitHub by using:

devtools::install_github("jinming-cheng/RankMap")

Quick Start

Example: predicting cell types for Xenium spatial data using a well-annotated single-cell reference dataset with RankMap.

library(RankMap)

# load data
seu_sc <- readRDS(system.file("extdata", "seu_sc.rds",
    package = "RankMap"
))

seu_xen <- readRDS(system.file("extdata", "seu_xen.rds",
    package = "RankMap"
))

# predict cell types
# if unsure about k (default 20), using k = 100 is a reasonable general choice
pred_df <- RankMap(
    ref_data = seu_sc,
    ref_labels = seu_sc$cell_type,
    new_data = seu_xen,
    k = 100
)

head(pred_df)

Citation

Please cite this article if you use RankMap:

To cite RankMap in publications, please use:

  Cheng J, Li S, Kim S, Ang C, Chew S, Chow P, Liu N (2026). "RankMap:
  Rank-based reference mapping for fast and robust cell type annotation
  in spatial and single-cell transcriptomics." _bioRxiv_.
  doi:10.64898/2026.03.01.708931
  <https://doi.org/10.64898/2026.03.01.708931>.

A BibTeX entry for LaTeX users is

  @Article{,
    title = {RankMap: Rank-based reference mapping for fast and robust cell type annotation in spatial and single-cell transcriptomics},
    author = {Jinming Cheng and Shengdi Li and Serim Kim and Chow Hiang Ang and Sin Chi Chew and Pierce Kah Hoe Chow and Nan Liu},
    journal = {bioRxiv},
    year = {2026},
    doi = {10.64898/2026.03.01.708931},
  }