Actually delete the file...

analysis/process-pbmc3k.Rmd

----
-title: "Process pbmc3k data"
-output: html_document
----
-
-```{r setup, include=FALSE}
-knitr::opts_chunk$set(echo = TRUE)
-```
-
-```{r libraries}
-library(TENxPBMCData)
-library(scater)
-library(scran)
-library(EnsDb.Hsapiens.v86)
-```
-
-```{r set-seed}
-set.seed(2332021)
-```
-
-```{r load-data}
-pbmc3k <- TENxPBMCData(dataset = "pbmc3k")
-counts(pbmc3k) <- as(counts(pbmc3k), "dgCMatrix")
-```
-
-```{r gene-annotation}
-rownames(pbmc3k) <- uniquifyFeatureNames(
-  rowData(pbmc3k)$ENSEMBL_ID, 
-  rowData(pbmc3k)$Symbol
-)
-
-# FIXME: Is this warning message safe to ignore?
-location <- mapIds(
-  EnsDb.Hsapiens.v86, 
-  keys = rowData(pbmc3k)$ENSEMBL_ID, 
-  column = "SEQNAME", keytype = "GENEID"
-)
-```
-
-The dataset is already filtered, so we don't perform cell detection.
-
-```{r quality-control}
-stats <- perCellQCMetrics(
-  pbmc3k, 
-  subsets = list(Mito = which(location == "MT"))
-)
-high_mito <- isOutlier(stats$subsets_Mito_percent, type = "higher")
-pbmc3k <- pbmc3k[, !high_mito]
-
-# Remove genes with 0 counts.
-zero_genes <- which(rowSums(counts(pbmc3k)) == 0)
-pbmc3k <- pbmc3k[-zero_genes, ]
-```
-
-```{r normalization}
-# Despite the name this is quite slow...
-clusters <- quickCluster(pbmc3k)
-pbmc3k <- computeSumFactors(pbmc3k, cluster = clusters)
-pbmc3k <- logNormCounts(pbmc3k)
-```
-
-```{r variance-modelling}
-dec_pbmc3k <- modelGeneVarByPoisson(pbmc3k)
-# Select the top 10000 genes.
-top_pbmc3k <- getTopHVGs(dec_pbmc3k, n = 10000)
-pbmc3k <- pbmc3k[top_pbmc3k, ]
-```
-
-```{r dimensionality-reduction}
-# Running PCA is slow.
-pbmc3k <- runPCA(pbmc3k)
-
-pbmc3k <- runTSNE(pbmc3k, dimred = "PCA")
-
-pbmc3k <- runUMAP(pbmc3k, dimred = "PCA")
-```
-
-```{r plot-tsne}
-plotTSNE(pbmc3k)
-```
-
-```{r plot-umap}
-plotUMAP(pbmc3k)
-```
-
-```{r clustering}
-# 15 is hand chose to give a clustering which 'looks right' on the tSNE plot.
-g <- buildSNNGraph(pbmc3k, k = 15, use.dimred = "PCA")
-clust <- igraph::cluster_walktrap(g)$membership
-colLabels(pbmc3k) <- factor(clust)
-```
-
-````{r plot-tsne-with-clustering}
-plotTSNE(pbmc3k, colour_by = "label")
-```
-
-```{r plot-umap-with-clustering}
-plotUMAP(pbmc3k, colour_by = "label")
-```
-
-In the simulations we only use a subset of cells to estimate the splatter parameters, so that the variance is not overestimated because of clustering.
-
-Here we select cluster 9. 
-
-```{r define-subset}
-subset_ind <- colLabels(pbmc3k) == 9
-pbmc3k_subset <- pbmc3k[, subset_ind]
-```
-
-```{r save-subset}
-saveRDS(pbmc3k_subset, here::here("data", "pbmc3k.rds"))
-```