Add rowData to pbmc3k and endothelial datasets

to fix Python indexing problems with Scanpy

code/prep-endothelial.R

@@ -20,6 +20,9 @@ endothelial@assays$HTO <- NULL
 endothelial@assays$SCT <- NULL
 
 endothelial <- as.SingleCellExperiment(endothelial)
+
+rowData(endothelial) <- DataFrame(gene = rownames(endothelial))
+
 colData(endothelial)$predicted.id <- factor(colData(endothelial)$predicted.id)
 colLabels(endothelial) <- colData(endothelial)$predicted.id
 

code/prep-pbmc3k.R

@@ -7,6 +7,8 @@ args <- R.utils::commandArgs(
   asValues = TRUE
 )
 
+.libPaths()
+
 suppressMessages({
   library(TENxPBMCData)
   library(scater)
@@ -23,6 +25,8 @@ rownames(pbmc3k) <- uniquifyFeatureNames(
   rowData(pbmc3k)$Symbol_TENx
 )
 
+old_rowdata <- rowData(pbmc3k)
+
 counts <- as(counts(pbmc3k), "dgCMatrix")
 colnames(counts) <- paste0("Cell", 1:ncol(counts))
 
@@ -81,6 +85,9 @@ pbmc3k <- RenameIdents(pbmc3k, new_cluster_ids)
 pbmc3k <- as.SingleCellExperiment(pbmc3k)
 colLabels(pbmc3k) <- colData(pbmc3k)$ident
 
+# Add back some rowData to prevent downstream (Python) errors.
+rowData(pbmc3k) <- rownames(pbmc3k)
+
 # Filter to the top 5000 genes.
 dec_pbmc3k <- modelGeneVarByPoisson(pbmc3k)
 top_pbmc3k <- getTopHVGs(dec_pbmc3k, n = 2000)