Data preparation for CBMC

Setup knitr and load utility functions

knitr::opts_chunk$set(echo = TRUE)
knitr::opts_knit$set(root.dir="E:/DISC/reproducibility")

utilities_path = "./source/utilities.r"
source(utilities_path)

Load Data

CBMC original data can be found here.
We filter cells following the original study.

gene_bc_mat = as.matrix(read.csv("./data/CBMC/original_data/GSE100866_CBMC_8K_13AB_10X-RNA_umi.csv.gz", sep = ",", strip.white = T, row.names = 1))
save_h5("./data/CBMC/original.loom", t(gene_bc_mat))

[1] TRUE

After read alignment and cell filtering, we assigned the species to each cell barcode. If more than 90% of UMI counts were coming from human genes, the cell barcode was considered to be human. If it was less than 10% of UMI counts, the assigned species was mouse. Cell barcodes in between 10% and 90% human were considered mixed species. The resulting assignment was 8,005 human, 579 mouse, 33 mixed.

human_genes = grep("human", tolower(rownames(gene_bc_mat)))
mouse_genes = grep("mouse", tolower(rownames(gene_bc_mat)))
human_library = colSums(gene_bc_mat[human_genes, ])
all_library = colSums(gene_bc_mat)
human_rate = human_library / all_library
human_mask = human_rate > 0.9
sum(human_mask)

[1] 8005

sum(human_rate < 0.1)

[1] 579

sum(human_rate <= 0.9 & human_rate >= 0.1)

[1] 33

Unless stated otherwise, analysis was performed on only the human cells and genes from the human reference genome.

gene_bc_filt = gene_bc_mat[human_genes, human_mask]
save_h5("./data/CBMC/raw.loom", t(gene_bc_filt))

[1] TRUE

Reference

Stoeckius, M. et al. Simultaneous epitope and transcriptome measurement in single cells. Nature methods 14, 865-868 (2017).