hvg fixes for summing

R/sum-duplicate-genes.R

@@ -76,8 +76,10 @@ sum_duplicate_genes <- function(sce, normalize = TRUE, recalculate_reduced_dims
     try({
       # try to cluster similar cells
       # clustering may fail if < 100 cells in dataset
-      qclust <- suppressWarnings(scran::quickCluster(summed_sce))
-      summed_sce <- scran::computeSumFactors(summed_sce, clusters = qclust)
+      suppressWarnings({
+        qclust <- scran::quickCluster(summed_sce)
+        summed_sce <- scran::computeSumFactors(summed_sce, clusters = qclust)
+      })
     })
     summed_sce <- scuttle::logNormCounts(summed_sce)
   }
@@ -90,8 +92,13 @@ sum_duplicate_genes <- function(sce, normalize = TRUE, recalculate_reduced_dims
       pca_dim <- min(50, ncol(sce) - 1) # always reduce dimensions at least 1!
     }
 
+    if (is.null(metadata(sce)$highly_variable_genes)) {
+      n_hvg <- 2000 # same default as ScPCA
+    } else {
+      n_hvg <- length(metadata(sce)$highly_variable_genes)
+    }
     hv_genes <- scran::modelGeneVar(summed_sce) |>
-      scran::getTopHVGs(n = length(metadata(sce)$highly_variable_genes))
+      scran::getTopHVGs(n = n_hvg)
     metadata(summed_sce)$highly_variable_genes <- hv_genes # store the new HVGs
 
     summed_sce <- scater::runPCA(summed_sce, subset_row = hv_genes, ncomponents = pca_dim) |>

tests/testthat/test-sum-duplicate-genes.R

@@ -8,6 +8,26 @@ test_that("merging works as expected", {
   expect_setequal(rownames(deduped_sce), rownames(sce))
 
 
+  expect_equal(
+    counts(deduped_sce)[rownames(sce), ], # need to make the row order identical
+    2 * counts(sce) |> Matrix::drop0() # original matrices may have explicit 0s
+  )
+  expect_equal(
+    assay(deduped_sce, "spliced")[rownames(sce), ],
+    2 * assay(sce, "spliced") |> Matrix::drop0()
+  )
+})
+
+test_that("merging works as expected with unprocessed SCE", {
+  sce <- readRDS(test_path("data", "filtered_sce.rds"))
+  double_sce <- rbind(sce, sce)
+  deduped_sce <- sum_duplicate_genes(double_sce)
+
+  expect_equal(dim(deduped_sce), dim(sce))
+
+  expect_setequal(rownames(deduped_sce), rownames(sce))
+
+
   expect_equal(
     counts(deduped_sce)[rownames(sce), ], # need to make the row order identical
     2 * counts(sce) |> Matrix::drop0() # original matrices may have explicit 0s