Skip to contents

Differential expression between two clusters or two groups of clusters.

Source: R/getMarkers.R
getDE.Rd

Pseudo-bulks cells by cluster via spe2PB and performs a two-sided quasi-likelihood test comparing the average of one set of clusters (cluster) against another (cluster_2), analogous to glmQLFTest. The contrast is the difference of the two group means, tested by fitting the null design with glmQLFit (see getMarkers for why the null design is used). The result table is post-processed like topTags.

Usage

getDE(
  spe,
  cluster,
  cluster_2,
  cluster_name = "cluster",
  dispersion = 0.05,
  method = c("quasi", "lrt"),
  n = Inf,
  adjust.method = "BH",
  sort.by = "PValue",
  p.value = 1,
  ...
)

Arguments

spe

A SpatialExperiment object.

cluster

Character vector. One or more cluster labels forming the first group. Positive logFC means up-regulated in this group.

cluster_2

Character vector. One or more cluster labels forming the second (reference) group. Must not overlap cluster.

cluster_name

Character. Column name in colData(spe) containing cluster labels. Default "cluster".

dispersion

Numeric. Fixed NB dispersion (BCV^2) used for the GLM fits. Default 0.05 (BCV \(\approx\) 0.224).

method

Character. Testing pipeline: "quasi" (default, glmQLFit quasi-deviance test) or "lrt" (glmFit + glmLRT). Both use the preset dispersion.

n

Integer. Maximum number of genes to return. Default Inf (all genes).

adjust.method

Character. Multiple-testing adjustment method passed to p.adjust. Default "BH".

sort.by

Character. How to order the table: "PValue" (default), "logFC", "stat", or "none".

p.value

Numeric. Cutoff on the adjusted p-value; genes above it are dropped. Default 1 (no filtering).

...

Additional arguments passed to glmQLFit for the null-model fit (used only when method = "quasi").

Value

A data frame with columns logFC (log2 fold change of cluster vs cluster_2), logCPM, Prop.1 and Prop.2 (proportion of cells expressing the gene in group 1 and group 2 respectively), stat (signed quasi statistic), PValue (two-sided), and the adjusted p-value column.

Examples


data("xenium_bc_spe")
spe <- normalizeAssay(spe)
spe <- runPCA(spe)
#> Genes with 0 variance are excluded: ENSG00000135218 NegControlProbe_00002 NegControlCodeword_0504 NegControlCodeword_0509 NegControlCodeword_0510 NegControlCodeword_0511 NegControlCodeword_0512 NegControlCodeword_0516 NegControlCodeword_0517 NegControlCodeword_0518 NegControlCodeword_0519 NegControlCodeword_0520 NegControlCodeword_0522 NegControlCodeword_0526 NegControlCodeword_0527 NegControlCodeword_0530 NegControlCodeword_0536 NegControlCodeword_0537 BLANK_0030 BLANK_0163 BLANK_0165 BLANK_0212 BLANK_0221 BLANK_0230 BLANK_0237 BLANK_0311 BLANK_0361 BLANK_0365 BLANK_0382 BLANK_0384 BLANK_0387 BLANK_0388 BLANK_0391 BLANK_0393 BLANK_0397 BLANK_0399 BLANK_0404 BLANK_0406 BLANK_0410 BLANK_0411 BLANK_0418 BLANK_0425 BLANK_0432 BLANK_0447
spe <- findNbrsSNN(spe, dimred = "PCA")
#> [1] "Getting K-nearest neighbour"
#> [1] "Getting shared nearest neighbour"
spe <- getClusters(spe, resolution = 0.5)
#> Reassigned all 1 cells from clusters with <= 1 cells (0 by graph, 1 by distance).
# cluster 1 vs cluster 2
de <- getDE(spe, cluster = 1, cluster_2 = 2)
# clusters 1 & 3 vs clusters 2 & 4
de2 <- getDE(spe, cluster = c(1, 3), cluster_2 = c(2, 4))