Confusion about simulating single-cell RNA data

[Peg-Wu]

Hi, Developer!

I would like to ask, when simulating single-cell RNA data, can GRN use the real human gene symbols instead of gene1, gene2, gene3,...

Thanks,
Peng

[lhc70000]

Hi,

I tested the following code and it worked:

grn <- data.frame(
  target = paste0("G", 2:20),
  source = paste0("G1"),
  weight = 1
)

res0 <- sim_true_counts(list2(
  num.cells = 500,
  num.genes = 30,
  num.cifs = 50,
  GRN = grn,
  cif.sigma = 0.1,
  tree = ape::read.tree(text = "(A:1,B:1,C:1,D:1);"),
  diff.cif.fraction = 0.8,
  do.velocity = F,
))

> rownames(res0$counts)
 [1] "G1"     "G10"    "G11"    "G12"    "G13"    "G14"    "G15"    "G16"    "G17"    "G18"    "G19"    "G2"    
[13] "G20"    "G3"     "G4"     "G5"     "G6"     "G7"     "G8"     "G9"     "gene21" "gene22" "gene23" "gene24"
[25] "gene25" "gene26" "gene27" "gene28" "gene29" "gene30"

However, I believe that there is a bug that will throw an error if num.genes is equal to the gene names listed in the GRN data frame (num.genes = 20 in the above example). I will fix this later, and in the meantime, you can modify the source code locally in 1_main.R Line 325:

  # name other genes
  L <- length(sim$gene_name_map)
  N_other <- sim$N$gene - L
+if (N_other > 0) {
    sim$gene_name_map <- c(
      sim$gene_name_map,
      setNames(seq(N_other) + L, paste0("gene", seq(N_other) + L))
    )
+}

and use devtools::load_all("/path/to/scMultiSim") to load the package.