Clean up clustermole_overlaps()

R/overlaps.R

@@ -18,17 +18,16 @@
 #' my_overlaps <- clustermole_overlaps(genes = my_genes, species = "hs")
 #' head(my_overlaps)
 clustermole_overlaps <- function(genes, species) {
-
   # check that the genes vector seems reasonable
   if (!is(genes, "character")) {
-    stop("genes is not a character vector")
+    stop("`genes` is not a character vector")
   }
-  genes <- unique(genes)
+  genes <- unique(sort(genes))
   if (length(genes) < 5) {
-    stop("too few genes")
+    stop("input should be at least 5 genes")
   }
   if (length(genes) > 5000) {
-    stop("too many genes")
+    stop("input should be less than 5,000 genes")
   }
 
   # retrieve markers
@@ -39,11 +38,13 @@ clustermole_overlaps <- function(genes, species) {
     dplyr::select(-dplyr::starts_with("gene")) %>%
     dplyr::distinct()
 
-  # check that input genes overlap species genes
+  # check that input genes overlap marker genes for a given species
   all_genes <- unique(markers_tbl$gene)
+  input_genes <- genes
   genes <- intersect(genes, all_genes)
-  if (length(genes) < 5) {
-    stop("the genes do not appear to correspond to the given species")
+  if (length(genes) < max(3, length(input_genes) * 0.2)) {
+    problematic_genes <- setdiff(input_genes, genes)
+    stop("large fraction of input genes are not known (possibly wrong species): ", toString(problematic_genes))
   }
 
   # run the enrichment analysis

tests/testthat/test-overlaps.R

@@ -6,10 +6,16 @@ markers_mm_tbl <- clustermole_markers(species = "mm")
 gene_names <- unique(markers_hs_tbl$gene)
 gene_names <- sample(gene_names)
 
+# generate strings that look like real genes
+fake_gene_names <- paste0(sample(LETTERS), sample(LETTERS), sample(LETTERS), sample(1:9, 1000, replace = TRUE))
+fake_gene_names <- sample(fake_gene_names)
+
 test_that("clustermole_overlaps() wrong input", {
   expect_error(clustermole_overlaps(gene_names[1:3], species = "hs"))
   expect_error(clustermole_overlaps(gene_names[1:10000], species = "hs"))
-  expect_error(clustermole_overlaps(c(gene_names[1:3], "X", "Y", "Z"), species = "hs"))
+  expect_error(clustermole_overlaps(fake_gene_names[1:100], species = "hs"))
+  expect_error(clustermole_overlaps(c(gene_names[1:2], fake_gene_names[1:5]), species = "hs"))
+  expect_error(clustermole_overlaps(c(gene_names[1:5], fake_gene_names[1:25]), species = "hs"))
   expect_error(clustermole_overlaps(as.list(gene_names[1:10]), species = "hs"))
 })
 
@@ -19,6 +25,12 @@ test_that("clustermole_overlaps() human input", {
   expect_gt(nrow(overlap_tbl), 1)
 })
 
+test_that("clustermole_overlaps() human input with fake genes", {
+  overlap_tbl <- clustermole_overlaps(genes = c(gene_names[1:10], fake_gene_names[1:20]), species = "hs")
+  expect_s3_class(overlap_tbl, "tbl_df")
+  expect_gt(nrow(overlap_tbl), 1)
+})
+
 # gene list for mouse overrepresentation tests
 gene_names <- unique(markers_mm_tbl$gene)
 gene_names <- sample(gene_names)