In shooshtarilab/TMExplorer: A Collection of Tumour Microenvironment Single-cell RNA Sequencing Datasets and Corresponding Metadata

knitr::opts_chunk$set(echo = TRUE)
library(TMExplorer)

Preparing Datasets For Upload

This is document is a brief example on how data can be prepared for upload to TMExplorer. I'll use GSE75688 since it has additional columns in the matrix and a separate cell-type info file.

First let's look at the dataset in TMExplorer.

gse75688 <- queryTME(geo_accession = 'GSE75688')[[1]]
dim(counts(gse75688))
counts(gse75688)[1:5,1:4]

The counts file is a 57915x563 matrix with a single column of rownames. This is the target format. Let's compare that to the dataset downloaded from GEO.

geo <- read.csv('GSE75688_GEO_processed_Breast_Cancer_raw_TPM_matrix.txt.gz',
                sep='\t')
dim(geo)
geo[1:5,1:4]

The GEO matrix has three additional columns. In order to match the target format, we'll drop the gene_type column, merge the gene_id and gene_name columns, then set the merged result as the rownames. NOTE: The rownames here can be gene names, gene IDs, some combination of the two like we have here, but they must be unique (R cannot set non-unique rownames) and identify the gene (instead of just being numbers).

geo$gene_type <- NULL
geo$gene_id <- paste(geo$gene_name, geo$gene_id,sep='_')
rownames(geo) <- geo$gene_id
geo$gene_id <- NULL
geo$gene_name <- NULL

dim(geo)
geo[1:5,1:4]
all(geo == counts(gse75688))

After modification, the dimensions are the same and all values in the matrix match the file in TMExplorer. If this was a new dataset, the modified geo object would be what you submit as the genes x cells matrix.

This dataset also has cell type information available. In order to prepare that for TMExplorer we'll need to load in the file and remove the extra columns.

geo_cellinfo <- read.csv('GSE75688_final_sample_information.txt',sep='\t')
head(geo_cellinfo)
geo_cellinfo$type <- NULL
geo_cellinfo$index <- NULL
geo_cellinfo$index2 <- NULL
names(geo_cellinfo)[2] <- "label"
dim(geo_cellinfo)
head(geo_cellinfo)

This file contains the cell type for each cell in the original dataset, as shown below.

table(geo_cellinfo$label)

shooshtarilab/TMExplorer documentation built on June 7, 2022, 1:31 p.m.