Package 'MultiK'

Pairwise SigClust tests between clusters

Description

Performs statistical significance testing for cluster separability using the SigClust method on all pairwise combinations of clusters.

Usage

CalcSigClust(seu, clusters, nsim = 100, nfeatures = 2000, cores = 1)

Arguments

seu	A Seurat object
clusters	Vector of cluster assignments
nsim	Number of simulations for SigClust. Default is 100
nfeatures	Number of variable features. Default is 2000
cores	Number of cores for parallel processing. Default is 1

Value

A symmetric matrix of p-values from pairwise SigClust tests. Diagonal elements are NA. P-values indicate significance of cluster separation.

---

Diagnostic plots for optimal K selection

Description

Generates three diagnostic plots from MultiK results: (1) Frequency distribution of K values (2) rPAC scores across K (3) Stability vs frequency trade-off plot

Usage

DiagMultiKPlot(ks, consensus)

Arguments

ks	Vector of K values from MultiK output
consensus	List of consensus matrices from MultiK output

Value

Combined ggplot object with 3 panels. Also prints optimal K to console.

---

Extract cluster assignments for specified K values

Description

Performs Seurat clustering and retrieves cluster assignments for the specified optimal K values.

Usage

getClusters(seu, optK, nPC = 30, k.param = 20, nfeatures = 2000)

Arguments

seu	A Seurat object
optK	Vector of target cluster numbers
nPC	Number of principal components. Default is 30
k.param	Number of nearest neighbors. Default is 20
nfeatures	Number of variable features. Default is 2000

Value

A list containing:

clusters	Matrix of cluster assignments (cells x K values)
resolution	Resolution parameters used for each K

---

MultiK: Determine optimal cluster numbers via consensus clustering

Description

Performs subsampling-based consensus clustering across multiple resolution parameters to identify robust cluster numbers in single-cell RNA-seq data.

Usage

MultiK(
  seu,
  resolution = seq(0.05, 2, 0.05),
  nPC = 30,
  reps = 100,
  pSample = 0.8,
  k.param = 20,
  nfeatures = 2000,
  seed = NULL,
  cores = 1
)

Arguments

seu	A Seurat object
resolution	Numeric vector of resolution parameters for Seurat clustering. Default is seq(0.05, 2, 0.05)
nPC	Number of principal components. Default is 30
reps	Number of subsampling iterations. Default is 100
pSample	Proportion of cells to subsample. Default is 0.8
k.param	Number of nearest neighbors for graph construction. Default is 20
nfeatures	Number of variable features to select. Default is 2000
seed	Random seed for reproducibility. Default is NULL
cores	Number of cores for parallel processing. Default is 1

Value

A list containing:

k	Vector of K values from all clustering runs
clusters	Clustering results for each K
consensus	Consensus matrix for each K

---

Three Cell Line Mixture Dataset

Description

A Seurat object containing single-cell RNA-seq data from a mixture of three cell lines (H2228, H1975, HCC827). This dataset is used for demonstrating the MultiK workflow.

Usage

p3cl

Format

A Seurat object with approximately 2,600 cells and the following structure:

assays

RNA assay with count and normalized data

meta.data

Cell metadata including cell line annotations

Source

Dong et al. 2019 (PMID: 31925417)

Examples

data(p3cl)
p3cl

---

Visualize SigClust results with dendrogram and heatmap

Description

Generates a combined visualization showing: (1) Hierarchical clustering dendrogram of cluster centroids (2) Heatmap of pairwise SigClust p-values

Usage

PlotSigClust(seu, clusters, pval, nfeatures = 2000)

Arguments

seu	A Seurat object
clusters	Vector of cluster assignments
pval	P-value matrix from CalcSigClust
nfeatures	Number of variable features. Default is 2000

Details

Node shapes indicate statistical significance: - Filled circles: significant separation (p < 0.05) - Open circles: non-significant separation (p >= 0.05)

Value

Invisibly returns NULL; displays the combined plot

---