getDAregion error: data are essentially constant

[stanaka6]

Hi DAseq team,

Thank you for providing a nice tool. I am having an error message and several warnings when running getDAregion. The error is probably related to the statistical test (t-test), but I am not sure if I can use other tests here, or something is wrong with my input data. My aim in performing DAseq is to see whether there are differentially abundant cell populations between WT and KO samples and identify if those DA populations are statistically and biologically reliable. My original Seurat object is an integrated object (WT: 2366 cells; KO: 5046 cells) created by the standard protocol.

Could you please help me to solve this issue? Any suggestions or comments would be appreciated.

Error and warnings:

da_regions_res001 <- getDAregion(
  X = integ,
  da.cells = da_cells,
  cell.labels = se@meta.data$orig.ident,
  labels.1 = label_res,
  labels.2 = label_nonres,
  resolution = 0.01,
  plot.embedding = embed
)

# Turning X to a matrix.
# Using min.cell = 50
# Warning: The following arguments are not used: row.names
# Warning: Feature names cannot have underscores ('_'), replacing with dashes ('-')
# Warning: The following arguments are not used: row.names
# Warning: Feature names cannot have underscores ('_'), replacing with dashes ('-')
# Warning: The following arguments are not used: row.names
# Warning: Feature names cannot have underscores ('_'), replacing with dashes ('-')
# Removing 3 DA regions with cells < 50.
# Error in t.test.default(x = idx.label.ratio[labels.2], idx.label.ratio[labels.1]) : 
#   data are essentially constant

The details of my data are below:

se <- readRDS("WT_KO_integrated_SeuratObject.rds")

se
# An object of class Seurat 
# 18075 features across 7412 samples within 2 assays 
# Active assay: integrated (2000 features, 2000 variable features)
#  1 other assay present: RNA
#  3 dimensional reductions calculated: pca, umap, tsne

integ <- data.frame(se@reductions$pca@cell.embeddings[,1:21])
label_res <- se$orig.ident[se$sample == "KO"]
label_nonres <- se$orig.ident[se$sample == "WT"]
embed <- data.frame(se@reductions$tsne@cell.embeddings)

head(embed)
#                           tSNE_1    tSNE_2
# WT_AAACCCAAGATCCGAG-1 -37.097319  11.12396
# WT_AAACCCACAGAGACTG-1 -16.785063  30.81720
# WT_AAACCCACATGAGAAT-1   1.463344  14.66381
# WT_AAACCCATCGAGAAAT-1  38.852505 -16.65050
# WT_AAACGAATCCTTGGAA-1  -7.666700  15.41251
# WT_AAAGAACAGCAACAAT-1  12.399879 -22.62338

tail(embed)
#                           tSNE_1     tSNE_2
# KO_TTTGTTGAGGTAAACT-1  34.348403 -12.031938
# KO_TTTGTTGCAGCGACCT-1 -22.358025 -18.574149
# KO_TTTGTTGCATACCATG-1   3.216129  10.567165
# KO_TTTGTTGCATATAGCC-1 -21.909047  24.071238
# KO_TTTGTTGGTTGGAGAC-1  -8.336157  -3.233347
# KO_TTTGTTGTCGCTTACC-1  -2.244519  43.394538

da_cells <- getDAcells(
  X = integ,
  cell.labels = se@meta.data$orig.ident,
  labels.1 = label_res,
  labels.2 = label_nonres,
  k.vector = seq(50, 500, 50),
  plot.embedding = embed
)

Thank you!

[klgoss]

I'm having the same issue. Has anyone found a solution?