scMetaTraj (version 0.1.1)

scMetaTraj_cluster: Cluster cells in metabolic PCA space

Description

scMetaTraj_cluster() identifies metabolic subclusters by constructing a kNN graph in metabolic PCA space and applying community detection.

IMPORTANT DESIGN PRINCIPLES:

• Clustering is performed ONLY in metabolic PCA space.

• UMAP coordinates must NEVER be used for clustering.

• Results are independent of transcriptomic clustering.

Usage

scMetaTraj_cluster(
  embedding,
  k = 20,
  resolution = 0.5,
  method = c("leiden", "louvain"),
  seed = 123
)

Value

A factor of length equal to number of cells, giving metabolic cluster labels per cell.

Arguments

embedding

Numeric matrix (cells x PCs). Output of scMetaTraj_embed(method = "PCA").

k

Integer. Number of nearest neighbors for kNN graph.

resolution

Numeric. Resolution parameter for clustering (used for Leiden only).

method

Character. "leiden" (default) or "louvain".

seed

Integer. Random seed for reproducibility.

Examples

# Create example PCA embedding
set.seed(123)
n_cells <- 100
n_pcs <- 5

embedding <- matrix(rnorm(n_cells * n_pcs), nrow = n_cells, ncol = n_pcs)
rownames(embedding) <- paste0("Cell", 1:n_cells)
colnames(embedding) <- paste0("PC", 1:n_pcs)

# Perform clustering
clusters <- scMetaTraj_cluster(
  embedding = embedding,
  k = 20,
  method = "louvain"
)

# View results
table(clusters)