SwitchClass is an R package for quantifying and visualising molecular reversibility and persistence across biological perturbations using a label-switch classification framework.
It provides a unified workflow for identifying baseline-aligned versus perturbation-aligned molecular features in longitudinal or comparative omics datasets.
The package was developed as part of the study Dissecting Reversal and Persistence of Molecular Features via a Label-Switch Classification Framework, which systematically maps molecular trajectories that normalise or remain dysregulated under different biological or therapeutic conditions.
SwitchClass implements the following core components:
-
Label-switch classification
A random-forest–based approach that compares importance profiles between two inverted label schemes to compute a per-feature delta score ((\delta = I_{rev} - I_{per})) indicating reversal versus persistence. -
Visualization utilities
Functions for visualizing molecular states via UMAP embeddings, quadrant-based scatterplots, feature-level boxplots, and annotated heatmaps. -
Downstream interpretation
Tools for pathway enrichment (Reactome) and quadrant-based biological annotation. -
Example datasets and vignettes
Includes demonstration data from colorectal cancer (CRC), phosphoproteomic, and immune-transcriptomic studies.
Install the development version from GitHub:
# install dependencies if not already installed
# BiocManager::install("PhosR")
# BiocManager::install("reactome.db")
# BiocManager::install("org.Hs.eg.db")
# BiocManager::install("annotate")
devtools::install_github("PYangLab/SwitchClass",
build_vignettes = TRUE,
dependencies = TRUE)
Please find our vignette:
browseVignettes("SwitchClass")If you have any enquiries about SwitchClass, please contact di.xiao@sydney.edu.au. We are also happy to receive any suggestions and comments.