FastCCCR

📖 Documentation: https://zaoqu-liu.github.io/FastCCCR/

Overview

FastCCCR is a high-performance R implementation of the FastCCC framework for cell-cell communication (CCC) inference in single-cell RNA sequencing (scRNA-seq) data. The package provides a rigorous statistical framework that computes exact null distributions via Fast Fourier Transform (FFT) convolution, integrates multiple statistical methodologies through Cauchy combination, and enables reference-based inference for rapid analysis of query datasets.

Methodological Framework

The FastCCCR analytical pipeline comprises five computational stages:

1. Single-Unit Expression Quantification: Aggregation of gene expression across cells within each cluster using configurable summary statistics (mean, median, or arbitrary quantiles)
2. Protein Complex Aggregation: Integration of multi-subunit protein complex expression via minimum or average aggregation functions
3. Ligand-Receptor Score Computation: Combination of ligand and receptor expression scores using arithmetic or geometric mean
4. Null Distribution Derivation: Exact computation of null distributions for cluster-level statistics using FFT-based convolution of single-cell PMFs
5. Multi-Method Integration: Cauchy combination of p-values from multiple statistical configurations to enhance robustness

Installation

R-universe (Recommended)

install.packages("FastCCCR", repos = "https://zaoqu-liu.r-universe.dev")

GitHub (Development Version)

# Install remotes if not available
if (!requireNamespace("remotes", quietly = TRUE))
    install.packages("remotes")

# Install FastCCCR from GitHub
remotes::install_github("Zaoqu-Liu/FastCCCR")

System Requirements

• R (≥ 4.1.0)
• C++ compiler with C++11 support (for Rcpp components)

Dependencies

Category	Packages
Core	R6, Rcpp, RcppArmadillo, data.table, Matrix
Parallel Computing	future, future.apply, progressr
Visualization	ggplot2, igraph, ComplexHeatmap
Single-Cell	Seurat (≥ 4.0)

Quick Start

Standard Analysis with Cauchy Combination

library(FastCCCR)
library(Seurat)

# Load single-cell data
seurat_obj <- readRDS("your_data.rds")

# Execute FastCCC with Cauchy combination (recommended)
results <- fastccc_cauchy(
    seurat_obj,
    database = "CPDBv5.0.0",
    celltype_col = "cell_type",
    min_percentile = 0.1,
    parallel = TRUE,
    n_cores = 4
)

# Extract significant interactions
significant_interactions <- results$significant_interactions
head(significant_interactions)

# Visualization
plot_ccc_network(results)
plot_ccc_heatmap(results)

Single Statistical Configuration

# Analysis with specified statistical method
results <- statistical_analysis_method(
    seurat_obj,
    database = "CPDBv5.0.0",
    celltype_col = "cell_type",
    single_unit_summary = "Mean",
    complex_aggregation = "Minimum",
    LR_combination = "Arithmetic",
    min_percentile = 0.1
)

Reference-Based Inference

Reference-based inference enables rapid CCC analysis by leveraging pre-computed reference panels from atlas-scale datasets.

# Step 1: Construct reference panel from atlas data
build_reference(
    seurat_obj = atlas_seurat,
    reference_name = "tissue_atlas",
    database = "CPDBv5.0.0",
    celltype_col = "cell_type",
    save_path = "./reference/"
)

# Step 2: Infer CCC in query dataset using reference
results <- infer_query(
    seurat_obj = query_seurat,
    reference_path = "./reference/tissue_atlas",
    database = "CPDBv5.0.0",
    celltype_col = "cell_type",
    celltype_mapping = list(
        "Reference_CellType1" = "Query_CellType1",
        "Reference_CellType2" = "Query_CellType2"
    )
)

Supported Databases

FastCCCR provides native support for multiple curated ligand-receptor interaction databases:

Database	Version	Species	Interactions
CellPhoneDB	v4.1.0	Human	~1,000
CellPhoneDB	v5.0.0	Human	~3,000
CellChat	v1.6.1	Human/Mouse	~2,000
NicheNet	v1.1.1	Human	~7,000
NicheNet	v2.1.5	Human	~7,000

Statistical Framework

Null Distribution Computation

FastCCCR computes exact null distributions for cluster-level expression statistics by modeling each cell's expression as an independent sample from the gene-specific empirical distribution. The distribution of the cluster mean (or quantile) is obtained through FFT-based convolution:

$$P(\bar{X}_n = k) = (F_X * F_X * \cdots * F_X)(k)$$

where $F_X$ denotes the single-cell PMF and $*$ represents convolution.

Cauchy Combination

Multiple p-values from different statistical configurations are combined using the Cauchy combination test:

$$T = \sum_{i=1}^{K} w_i \tan\left(\pi(0.5 - p_i)\right)$$

$$p_{\text{combined}} = 0.5 - \frac{\arctan(T)}{\pi}$$

This approach provides robust inference while controlling the false discovery rate.

Performance Considerations

• Vectorized Operations: Core computations leverage data.table and Matrix for efficient memory usage
• C++ Backend: FFT convolution implemented via RcppArmadillo for computational efficiency
• Parallel Processing: Native support for parallel computation via the future framework
• Memory Efficiency: Sparse matrix operations for large-scale datasets

Documentation

• Package Documentation
• API Reference

Citation

If you utilize FastCCCR in your research, please cite:

@software{liu2026fastcccr,
  author = {Liu, Zaoqu},
  title = {FastCCCR: Fast Cell-Cell Communication Analysis with Statistical Framework},
  year = {2026},
  url = {https://github.com/Zaoqu-Liu/FastCCCR}
}

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contact

• Author: Zaoqu Liu
• Email: liuzaoqu@163.com
• GitHub: https://github.com/Zaoqu-Liu
• Issues: GitHub Issues