v2.5.0

Highlights

• Streamlined code-base – major internal refactor for clarity, speed and a ~20 % smaller dependency tree.
• Consistent, flexible visualisation API across all plotting helpers.
• Robust unit-test suite (>250 expectations) now ships with the package.

New & enhanced functionality

Area	Function(s)	What changed
Visualisation	ridgeEnrichment()	True gradient coloring mode for numeric color.by; optional per-cell rugs; quantile median line; fixed grey-fill bug
	densityEnrichment()	accepts new rug.height; ~4× faster ranking routine using MatrixGenerics::rowMeans2; cleaner two-panel layout via patchwork
	gseaEnrichment()	new rug.height; clearer legend showing ES/NES/ p; internal vectorised ES calculation
	splitEnrichment()	rewritten: split violins when split.by has 2 levels, dodged violins otherwise; inline boxplots; auto Z-scaling; palette helper
	scatterEnrichment()	density-aware points (via ggpointdensity), hex-bin alternative, optional Pearson/Spearman overlay, continuous or discrete color mapping
Dimensionality reduction	performPCA() / pcaEnrichment()	uses irlba::prcomp_irlba() automatically for large matrices; stores eigen-values/contribution in misc; add.percent.contribution now always respected
Scoring backend	escape.matrix() / .compute_enrichment()	lazy loading of heavy back-ends (GSVA, UCell, AUCell); unified .build_gsva_param(); drops empty gene-sets up-front
Normalization	performNormalization()	chunk-wise expressed-gene scaling (memory-friendly); accepts external scale.factor; optional signed log-transform; returns object with assay <assay>_normalized
Gene-set retrieval	getGeneSets()	downloads now cached under tools::R_user_dir("escape", "cache"); graceful KEGG append; clearer error for non-human/mouse requests

Performance & dependency reductions

• Replaced plyr, stringr, rlang usage with base-R helpers; these packages
  are now Suggests only.
• Common color and label utilities (.colorizer(), .colorby(), .orderFunction())
  removed redundant tidyverse imports.
• Internal matrices split/chunked with new .split_* helpers to cap memory
  during parallel scoring/normalization.

Bug fixes

• Gradient mode in ridgeEnrichment() no longer produces grey fills when the
  chosen gene-set is mapped to color.by.
• pcaEnrichment() axis labels correctly include variance contribution
  when display.factors = FALSE.
• .grabDimRed() handles both Seurat v5 and <v5 slot structures; fixes missing
  eigen-values for SCE objects.
• escape.matrix() respects min.size = NULL (no filtering) and handles
  zero-overlap gene-sets gracefully.
• Global variable declarations consolidated – eliminates R CMD check NOTES
  regarding na.omit, value, etc.

Documentation

• DESCRIPTION rewritten – heavy packages moved to Suggests; added explicit
  Config/reticulate for BiocParallel.
• escape.gene.sets data object now fully documented with source, usage, and reference.