In Jerby-Lab/opipes: What the Package Does (One Line, Title Case)

knitr::opts_chunk$set(echo = TRUE)

Load opipes, the Jerby Lab's omics package

library(opipes)

Load data

Here we load one sample of single cell MERFISH data on an ovarian cancer tumor

out <- readRDS("~/Downloads/20220920_sample3.rds")
tpm <- do.call("cbind", out[["counts"]])
meta <- do.call("rbind", out[["meta"]])
row.names(meta) <- meta$cellid

Subsample the data

If your dataset is very large and you need to subsample data, then you can use this functionality below. Here we set p=1 because we want to use all the data.

sub <- subsample(tpm, meta, p =1, replace = F)
tpm <- sub[["c"]]
meta <- sub[["m"]]

Seurat: Dimension Reduction and Clustering

This package includes some standard Seurat functions for getting started on the bread and butter steps in processing single cell RNA-seq data. Note that the UMAP step usually takes a while.

so <- seuratify(tpm, prj_name = "tutorial_data_set") # create Seurat Object
so <- std_preprocess(so) # normalize and scale
so <- run_pca(so, n_pcs=10) # run PCA
so <- umap_tsne(so, umap.flag = T, n_dims = 10) # run UMAP
so <- cluster(so, n_dims = 10, reduction = "pca") # cluster (sNN)

Cell-type Assignments

We have our own in-house cell-type assignment algorithm. This is how to use it:

# create r object 
r <- c()
r$tpm <- tpm 
r$genes <- row.names(tpm)
r$cells <- colnames(tpm)
r$clusters <- paste0("C", so$RNA_snn_res.0.4)

# read in your preferred cell type signatures
cell.sig <- readRDS("~/Downloads/cell.type.sig.rds")
cell.sig <- cell.sig$HGSC[-c(3,4,10)] # removing CD8.T, CD4.T, and cell cycle genes for now

# execute to cell-type assignment code
r <- scRNA_markers_assign(r = r, 
                     cell.clusters = r$clusters,
                     cell.sig = cell.sig,
                     non.immune.cell.types = c('Endothelial','CAF','Malignant'),
                     immune.markers = c("CD3E","PTPRC","CD8A" , "CD4"))

Now let's plot our results

First we make a data-frame with all the data that you plan on plotting, including the UMAP dimensions, cell-types, clusters, etc.

library(ggplot2)
plot <- data.frame(so@reductions$[e-mail], # UMAP
                   cluster = so$RNA_snn_res.0.4, # clusters
                   cell.types = r$cell.types) # cell types

And here are the fruits of your labor:

# plot UMAP with clusters
ggplot(plot, aes(x = UMAP_1, y = UMAP_2, col = cluster)) + 
  geom_point(size = 0.2) +
  theme_bw() + 
  guides(colour = guide_legend("Cluster", override.aes = list(size=3, shape = 15))) + 
  theme(plot.title = element_text(hjust = 0.5, size = 18, face = "bold"))  

ggplot(plot, aes(x = UMAP_1, y = UMAP_2, col = cell.types)) + 
  geom_point(size = 0.2) +
  theme_bw() + 
  guides(colour = guide_legend("Cluster", override.aes = list(size=3, shape = 15))) + 
  theme(plot.title = element_text(hjust = 0.5, size = 18, face = "bold"))  

saveRDS(so, file = "../../hgosc/intermediates/20220323_precomputed.rds")
saveRDS(plot, file = "../../hgosc/intermediates/20220323_precomputed_plot.rds")

Jerby-Lab/opipes documentation built on Oct. 7, 2022, 12:28 p.m.