scSLIDE

Single-cell Sample-Level Integration using Density Estimation

scSLIDE is an R package to perform sample-level analysis for multi-sample single-cell RNA sequencing data. It leverages a semi-supervised dimensional reduction framework to embed cells into a latent space that robustly retains both their underlying type- and state-identity as well as phenotype-driven differences. Each sample is then represented as a probability distribution of cellular states, yielding a sample-level representation that can be directly used for clustering, trajectory inference, and integrative analyses.

Key Features

Dimensionality Reduction

  • RunPLS: Partial Least Squares (PLS) dimensionality reduction with support for plsr, spls, and cppls methods
  • RunDiffusionMap: Diffusion map analysis for trajectory inference

Sample-Level Analysis

  • PrepareSampleObject: Comprehensive workflow for preparing single-cell data for sample-level analysis
  • GenerateSampleObject: Generate sample-level count matrices from single-cell data

Novel Differential Expression Test

  • TrajDETest: Trajectory-based differential expression analysis using negative binomial regression

Installation 

# Install BiocManager if not already installed
if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

# Install smoother which is required by "destiny" but recently got archived. 
install.packages("https://cran.r-project.org/src/contrib/Archive/smoother/smoother_1.1.tar.gz", repos = NULL, type = "source")

# Install Bioconductor dependencies manually
BiocManager::install(c("glmGamPoi", "destiny"))

# Install SeuratObject 
install.packages("SeuratObject")

# Install Seurat from a developmental branch that is compatible with scSLIDE (built upon v5.3.1) 
remotes::install_github("satijalab/seurat", "v5.3.1_scSLIDE_compatible")

# Install scSLIDE from GitHub
devtools::install_github("satijalab/scSLIDE")

Dependencies

scSLIDE depends on: - Seurat (>= 5.3.1): Core single-cell analysis framework - SeuratObject (>= 5.2.0): Seurat object structure - pls: Partial least squares regression - spls: Sparse partial least squares - glmGamPoi: Gamma-Poisson regression for DE analysis - destiny: Diffusion map analysis - ggplot2, dplyr, tidyr, RColorBrewer: Visualization and data manipulation

Tutorial

A tutorial for running the scSLIDE package is provided here.

Citation

If you use scSLIDE in your research, please cite:

https://doi.org/10.64898/2025.12.10.693462

License

MIT License - see LICENSE file for details.

Support

For questions and support, please open an issue on the GitHub repository.