Package 'SwitchClass'

Assign SwitchClass quadrants

Description

Q1: delta > T & FC > 0 Q2: delta < -T & FC > 0 Q3: delta < -T & FC < 0 Q4: delta > T & FC < 0 Q0: otherwise

Usage

assign_quadrants(delta, fc, delta_thresh = 0.15)

Arguments

delta	named numeric vector (features).
fc	named numeric vector (same feature names as delta).
delta_thresh	numeric threshold for |delta|.

Value

factor vector of quadrant labels with names = features.

Examples

delta <- c(a = 0.2, b = -0.2, c = -0.2, d = 0.2, e = 0)
fc <- c(a = 1, b = 1, c = -1, d = -1, e = 1)
assign_quadrants(delta, fc, delta_thresh = 0.15)

---

CRC subset data

Description

A small example dataset from the colorectal cancer (CRC) plasma proteome study. It contains a pre-filtered expression matrix and sample group labels (HC = healthy control, Pre = pre-treatment, Post = post-treatment).

Usage

data(CRC_subset)

Format

A list with two elements:

expr

A numeric matrix of features (rows) × samples (columns).

group

A character vector of sample groups, length = ncol(expr).

Details

The data are a subset of the discovery cohort used in the SwitchClass analysis. Useful for testing or demonstrating functions such as label_switch_classify().

Source

Internal preprocessed subset from Li et al. 2024, Nat Commun.

Examples

data(CRC_subset)
dim(CRC_subset$expr)
table(CRC_subset$group)

---

Reactome ORA per quadrant

Description

Wraps a convenience pipeline:

• builds Reactome TERM2GENE via msigdbr

• runs pathwayOverrepresent on each quadrant (Q1..Q4)

• returns top-k barplots (-log10 p) and tidy tables

Usage

enrich_reactome_quadrants(
  df,
  universe,
  species = "Homo sapiens",
  top_terms = 5,
  palette = c(Q1 = "#33DBB2", Q2 = "#DFB647", Q3 = "#2ECEF8", Q4 = "#76CF49")
)

Arguments

df	data.frame with columns feature and quadrant (Q1..Q4). If quadrant is absent, columns delta and fc are used to assign it by sign.
universe	character vector of background gene symbols (e.g., rownames(discovery.cohort))
species	msigdbr species (default "Homo sapiens")
top_terms	how many top terms per quadrant to plot (default 5)
palette	named colors for bars of Q1..Q4

Value

list with:

results

named list of data.frames p1..p4 (or NULL if no hits)

barplot_df

tidy data.frame used for plotting

plot

ggpubr patch of 4 panels

---

Label-switch classification (with optional multicore)

Description

Trains two random-forest classifiers under baseline-aligned and perturbation-aligned label schemes, then returns the directional importance score $\delta = I_{baseline} - I_{perturbation}$ using MeanDecreaseGini importances.

Usage

label_switch_classify(
  X,
  y_reverse,
  y_persist,
  ntimes = 25,
  seed = 1,
  ntree = 1000,
  mtry = NULL,
  parallel = FALSE,
  ncores = parallel::detectCores() - 1L
)

Arguments

X	numeric matrix with features in rows and samples in columns.
y_reverse	character/factor labels (length = ncol(X)) for the baseline-aligned label scheme. The argument name is retained for compatibility with earlier package versions.
y_persist	character/factor labels (length = ncol(X)) for the perturbation-aligned label scheme. The argument name is retained for compatibility with earlier package versions.
ntimes	integer; number of RF repeats to average (default 25).
seed	integer; base seed (default 1).
ntree	integer; trees per forest (default 1000).
mtry	integer or NULL; variables tried at each split (default NULL).
parallel	logical; use multicore (default FALSE).
ncores	integer; cores if parallel=TRUE (default detectCores()-1).

Value

list with named numeric vectors: delta, importance_baseline, importance_perturbation, plus compatibility aliases I_rev and I_per.

Examples

X <- matrix(seq_len(60) / 10, nrow = 10)
rownames(X) <- paste0("feature", seq_len(nrow(X)))
y_baseline <- rep(c("c1", "c2", "c1"), each = 2)
y_perturbation <- rep(c("c1", "c2", "c2"), each = 2)
res <- label_switch_classify(
  X, y_baseline, y_perturbation, ntimes = 1, ntree = 5
)
head(res$delta)

---

Scatter plot of delta vs fold-change with quadrant assignment

Description

Builds a tidy data.frame with quadrants and returns the ggplot scatter.

Usage

plot_delta_vs_fc(
  delta,
  fc,
  delta_thresh = 0.15,
  label_top = 5,
  label_size = 3,
  palette = c(Q1 = "#2E8B57", Q2 = "#C0392B", Q3 = "#C0392B", Q4 = "#2E8B57", Q0 =
    "grey70"),
  point_size = 2.2,
  alpha = 0.85,
  xlab = expression(delta),
  ylab = expression(log[2] * " FC"),
  title = "feature importance vs log2 fold-change"
)

Arguments

delta	named numeric vector of directional importance scores (features).
fc	named numeric vector of log2 fold-change (same names as delta).
delta_thresh	numeric; absolute delta threshold to define Q1-Q4 (default 0.15).
label_top	integer; number of labels per quadrant (by absolute delta). Set 0 to skip.
label_size	numeric; text size for labels.
palette	named colors for quadrants. Names should include Q1..Q4 and optional Q0.
point_size	numeric; scatter point size.
alpha	numeric from 0 to 1; point transparency.
xlab, ylab	axis labels.
title	plot title.

Value

list with elements: data (data.frame) and plot (ggplot object).

Examples

delta <- c(a = 0.2, b = -0.2, c = -0.2, d = 0.2, e = 0)
fc <- c(a = 1, b = 1, c = -1, d = -1, e = 1)
out <- plot_delta_vs_fc(delta, fc, label_top = 0)
head(out$data)
out$plot

---

Boxplots for selected features across groups

Description

Boxplots for selected features across groups

Usage

plot_feature_boxplots(
  X,
  group,
  features,
  feature_order = NULL,
  palette = NULL,
  ncol = 4,
  add_jitter = TRUE,
  jitter_width = 0.15,
  notch = FALSE,
  free_y = TRUE,
  point_size = 1.6,
  box_linewidth = 0.4,
  title = NULL,
  ylab = "log2 expression",
  xlab = NULL,
  save_path = NULL,
  width = 7,
  height = 4.5
)

Arguments

X	numeric matrix with features in rows and samples in columns.
group	factor/character of length ncol(X) giving sample groups.
features	character vector of feature (gene/protein) names to plot.
feature_order	optional character vector to set facet order (defaults to features order).
palette	optional named vector of group colors; if NULL, generated automatically.
ncol	integer; facets per row (default 4).
add_jitter	logical; add jittered points (default TRUE).
jitter_width	numeric; jitter width (default 0.15).
notch	logical; use notched boxplots (default FALSE).
free_y	logical; facet with free y scales (default TRUE).
point_size	numeric; jitter point size (default 1.6).
box_linewidth	numeric; box outline width (default 0.4).
title, ylab, xlab	plot labels.
save_path	optional file path to save (pdf/png) using ggplot2::ggsave.
width, height	numeric dimensions (inches) if saving.

Value

ggplot object.

Examples

X <- matrix(seq_len(40) / 10, nrow = 5)
rownames(X) <- paste0("feature", seq_len(nrow(X)))
group <- rep(c("A", "B"), each = 4)
plot_feature_boxplots(X, group, features = rownames(X)[1:2], ncol = 2)

---

Plot heatmap for quadrant-classified features

Description

Generates a standardized heatmap of feature expression grouped by quadrant (Q1-Q4) and sample group (e.g., HC / Pre / Post). Quadrants are optionally filtered by delta cutoff thresholds.

Usage

plot_heatmap_quadrants(
  X,
  group,
  df_quad,
  cutoff = 0.15,
  palette_group = NULL,
  palette_quad = NULL,
  scale = TRUE,
  cap = 2,
  show_rownames = TRUE,
  show_colnames = TRUE,
  cluster_rows = FALSE,
  cluster_cols = FALSE
)

Arguments

X	numeric matrix with features in rows and samples in columns.
group	factor/character of length = ncol(X); sample group labels.
df_quad	data.frame with columns: feature, quadrant, and delta.
cutoff	numeric; delta threshold for selecting strong quadrant features (default 0.15). Quadrants Q1 and Q4 use delta > cutoff; Q2 and Q3 use delta < -cutoff.
palette_group	optional named color vector for sample groups.
palette_quad	optional named color vector for quadrants.
scale	logical; z-score features (default TRUE).
cap	numeric; cap scaled values at +/- cap (default 2).
show_rownames, show_colnames	logical; show labels (default TRUE).
cluster_rows, cluster_cols	logical; cluster rows/columns (default FALSE).

Value

Invisibly returns the pheatmap object.

Examples

plot_heatmap_quadrants(
  X = matrix(
    seq_len(24), nrow = 4,
    dimnames = list(paste0("feature", 1:4), paste0("sample", 1:6))
  ),
  group = rep(c("A", "B"), each = 3),
  df_quad = data.frame(
    feature = paste0("feature", 1:4),
    quadrant = c("Q1", "Q2", "Q3", "Q4"),
    delta = c(0.2, -0.2, -0.2, 0.2)
  ),
  show_rownames = FALSE,
  show_colnames = FALSE
)

---

Quick ggplot for UMAP embedding

Description

Quick ggplot for UMAP embedding

Usage

plot_umap(
  emb,
  groups,
  title = "",
  point_size = 3,
  shape = 21,
  color = "black",
  stroke = 1.1,
  alpha = 0.9,
  palette = NULL
)

Arguments

emb	matrix from run_umap_samples() (samples x 2).
groups	vector of sample groups (length = nrow(emb)).
title	character; plot title.
point_size	numeric; point size (default 3).
shape	integer; point shape (default 21 = filled circle).
color, stroke, alpha	aesthetics for outlines and transparency.
palette	optional named vector of fill colors; if NULL, generated automatically.

Value

ggplot object.

Examples

emb <- matrix(c(0, 0, 1, 1, 0, 1), ncol = 2, byrow = TRUE)
rownames(emb) <- paste0("sample", seq_len(nrow(emb)))
plot_umap(emb, groups = c("A", "A", "B"))

---

UMAP on samples using selected features

Description

UMAP on samples using selected features

Usage

run_umap_samples(
  X,
  feats,
  n_neighbors = 15,
  min_dist = 0.1,
  metric = "euclidean",
  seed = 1
)

Arguments

X	numeric matrix with features in rows and samples in columns.
feats	character; feature names to use (non-matching are ignored with a message).
n_neighbors	integer; UMAP neighbor size (default 15).
min_dist	numeric; UMAP min_dist (default 0.1).
metric	character; distance metric (default "euclidean").
seed	integer; RNG seed for reproducibility (default 1).

Value

matrix with samples in rows and two UMAP columns, with rownames = sample IDs.

Examples

X <- matrix(seq_len(80) / 10, nrow = 10)
rownames(X) <- paste0("feature", seq_len(nrow(X)))
colnames(X) <- paste0("sample", seq_len(ncol(X)))
emb <- run_umap_samples(X, rownames(X)[1:4], n_neighbors = 3)
head(emb)

---

UMAP triptych using delta-ranked features

Description

Generates UMAPs for (1) baseline-aligned (+delta), (2) perturbation-aligned (-delta), and (3) all features. Returns the three embeddings and a combined plot.

Usage

umap_triptych_by_delta(
  X,
  delta,
  groups,
  top_n = 100,
  n_neighbors = 15,
  min_dist = 0.1,
  metric = "euclidean",
  seed = 1,
  palette = NULL
)

Arguments

X	numeric matrix with features in rows and samples in columns.
delta	named numeric vector of delta scores (names = feature IDs).
groups	factor or character of sample groups (length = ncol(X)).
top_n	integer; number of features to use per side (default 100).
n_neighbors, min_dist, metric, seed	passed to run_umap_samples().
palette	optional named color vector for groups; if NULL, generated automatically.

Value

list with emb_baseline, emb_perturbation, emb_all, and plot.

Examples

X <- matrix(seq_len(100) / 10, nrow = 10)
rownames(X) <- paste0("feature", seq_len(nrow(X)))
colnames(X) <- paste0("sample", seq_len(ncol(X)))
delta <- stats::setNames(seq(-1, 1, length.out = nrow(X)), rownames(X))
groups <- rep(c("A", "B"), each = 5)
triptych <- umap_triptych_by_delta(
  X, delta, groups, top_n = 2, n_neighbors = 3
)
names(triptych)

---