RaCInG (Random Cell-cell Interaction Generator) reconstructs patient-specific cell-cell communication networks from bulk RNA-seq data and extracts network-level features using either a kernel-based or Monte Carlo workflow.
This package is the R implementation of the original RaCInG Python framework, as described in van Santvoort et al. (2025), repackaged for seamless integration with R/Bioconductor analysis pipelines. It is additionally coupled with multideconv for deconvolution and cell-type subgroup identification, and with OmnipathR/liana for ligand-receptor prior knowledge.
# install.packages("remotes")
remotes::install_github("VeraPancaldiLab/RaCInG_package")
library(RaCInG)# install.packages("devtools")
devtools::install(".")If you want to start directly from raw counts with
prepare_input_files(), install the optional helper packages used
during deconvolution and prior-network construction:
install.packages(c("ggplot2", "nnls"))
# ADImpute and OmnipathR are available from Bioconductor:
BiocManager::install(c("ADImpute", "OmnipathR"))
# liana and multideconv are GitHub-only:
remotes::install_github(c("saezlab/liana", "VeraPancaldiLab/multideconv"))| Goal | Function | Output |
|---|---|---|
| Build input matrices from raw counts | prepare_input_files() |
Named list with L, R, C, LR matrices and labels |
| Run deterministic features | compute_racing_kernel() |
Kernel arrays + feature matrix |
| Run simulation-based features | compute_racing_montecarlo() |
Monte Carlo summaries |
| Compare patient groups | wilcox_group_test() |
Statistics table for downstream plots |
library(RaCInG)
# Build input matrices from a real bulk RNA-seq counts matrix
# (bundled with the optional `multideconv` dependency)
data(raw_counts, package = "multideconv")
counts_matrix <- as.matrix(raw_counts)
input <- prepare_input_files(
counts = counts_matrix,
output_folder = "Results/",
file_name = "example"
)
# Run kernel method (from raw counts)
kernel_res <- compute_racing_kernel(
counts = counts_matrix,
file_name = "example",
output_folder = tempdir(),
communication_type = "W"
)
# Or pass pre-computed inputs to skip preprocessing
kernel_res <- compute_racing_kernel(
input_data = input,
communication_type = "W"
)
# Monte Carlo method
mc_res <- compute_racing_montecarlo(
input_data = input,
file_name = "example",
output_folder = tempdir(),
communication_type = "W",
Ncells = 100,
Ngraphs = 10,
Ndegree = 3
)
# Compare clinical groups or correlate features with a continuous score
grouping <- c("Responder", "Responder", "Non-responder", "Non-responder")
wilcox_results <- wilcox_group_test(kernel_res$features, grouping)
top_features_plot(wilcox_results)Both compute_racing_kernel() and compute_racing_montecarlo() also accept
communication_type as a vector (e.g. c("D", "W", "TT", "GSCC")), extracting every
requested feature family from the same computed kernel / same simulated graphs in one
call, instead of repeating the expensive step per type. compute_racing_montecarlo()
additionally takes ncores to parallelize independent patients across cores.
- 📘 Vignette: Getting started with RaCInG
- 🌐 Website: https://VeraPancaldiLab.github.io/RaCInG_package/
- 🐍 Original Python implementation: https://github.com/SysBioOncology/RaCInG
If you use this package, please cite the RaCInG publication:
van Santvoort M, Lapuente-Santana Ó, Zopoglou M, Zackl C, Finotello F, van der Hoorn P & Eduati F (2025). Mathematically mapping the network of cells in the tumor microenvironment. Cell Reports Methods, 5(2), 100985.
This R package implementation was developed by Marcelo Hurtado from the Pancaldi team, led by Vera Pancaldi. Marcelo is currently the primary maintainer of the package.