getSpatiallyInformedClusters: Get spatially informed clusters by weighting graph-based...

In JEFworks/MERingue: Characterizing spatial gene expression heterogeneity in spatially resolved single-cell transcriptomics data with non-uniform cellular densities

Get spatially informed clusters by weighting graph-based clustering with spatial information

Description

Get spatially informed clusters by weighting graph-based clustering with spatial information

Usage

getSpatiallyInformedClusters(
  pcs,
  W,
  k,
  alpha = 1,
  beta = 1,
  method = igraph::cluster_louvain,
  verbose = FALSE,
  details = FALSE
)

Arguments

pcs	A matrix of principal components or gene expression to assess transcriptional similarity
W	Binary adjacency matrix
k	Number of nearest neighbors for clustering
alpha	Pseuodocount in edge weight 1/(alpha + as.vector(pweight)) + beta (default: 1)
beta	Pseudocount in edge weight 1/(alpha + as.vector(pweight)) + beta (default: 1)
method	igraph method for graph-based clustering (default: cluster_louvain)
verbose	Verbosity
details	Return detailed ouputs

Value

Factor of cluster annotations

Examples

# simulate 3 spatially but 2 transcriptionally distinct groups
N <- 300
M <- 30
# Three spatially distinct groups
pos1 <- cbind(rnorm(N/3), rnorm(N/3))
pos2 <- cbind(rnorm(N/3, 10), rnorm(N/3))
pos3 <- cbind(rnorm(N/3, 10), rnorm(N/3, 10))
pos <- rbind(rbind(pos1, pos2), pos3)
group <- c(rep(1, N/3), rep(2, N/3), rep(3, N/3))
names(group) <- rownames(pos) <- paste0('cell', 1:N)
# But two are transcriptionally identical
pcs12 <- matrix(rnorm(N*2/3*M), N*2/3, M)
pcs3 <- matrix(rnorm(N*1/3*M, 10), N*1/3, M)
pcs <- rbind(pcs12, pcs3)
pcs <- cbind(pcs, abs(10-pcs))
colnames(pcs) <- paste0('PC:', 1:ncol(pcs))
rownames(pcs) <- rownames(pos)
W <- getSpatialNeighbors(pos, filterDist=5)
com <- getSpatiallyInformedClusters(pcs, W, k=50, verbose=TRUE)

JEFworks/MERingue documentation built on June 11, 2022, 4:16 a.m.