scCompoundDE

Compositional and Transcriptional Decomposition of Pseudo-Bulk Differential Expression

Bioconductor version: 3.24 · Package version: 0.99.1

scCompoundDE decomposes pseudo-bulk differential expression (DE) signals into two orthogonal components: transcriptional changes (cell-intrinsic expression shifts) and compositional changes (shifts in the relative abundance of cell subtypes). Standard pseudo-bulk DE tools confound these two sources of signal, producing spurious DE calls when subtype proportions differ between conditions. scCompoundDE fits per-subtype limma-voom models, estimates subtype proportion shifts, and uses a z-score-normalized decomposition to assign each gene a TC_ratio score — the fraction of its DE signal attributable to transcription versus composition. Genes are then classified as transcriptional (real biology), compositional (artifact), or mixed (requires caution). All functions operate natively on SingleCellExperiment objects and return a CDEResult S4 object that extends the standard DE output with full decomposition statistics.

if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install("scCompoundDE")