Hello,
I am trying to convert the digital expression matrix of a single cell RNA-seq dataset on GEO to a counts matrix (e.g. rows = genes, columns = samples, data = gene counts) and was wondering if anyone knew how to do that.
Here is the link to the dataset:
https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE133486
Thank you.
Try the function Matrix::readMM from the Matrix package in R. It reads the MatrixMarket (mtx) files which are the counts. The barcodes, features etc you can read with any reader function, e.g. read.delim.
You will then have to assemble the data pbject you want, so putting features as rownames and barcodes as colnames. There are container formats for single-cell data, e.g. the SingleCellExperiment class in Bioconductor that might be worth reading about, which can also store dimensionality reductions, metadata etc.
Seconding this, it really comes down to (a) reading in the mtx file as a sparse matrix, which often only contains the count values which is why you also need to read in the gene names (--> features) and barcodes/cell names. And the metadata will probably just be a data.frame'able table, too, with additional information about the cells.
If you want even more convenience, you could
read10xCounts() function from the DropletUtils package. This will create a SingleCellExperiment object.You are right, not all GEO submissions may have mtx files, but the one that is linked above does, e.g. GSE133486_10XAdiposeNuclei.digital_expression.mtx.gz.
If there's no mtx, there's usually just another plain text file containing the counts, which can be read into R with the usual functions that ATpoint mentioned.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
@Friederike Thank you
PS
Remark: examples how to read "mtx" files by R and Python can be found here: https://www.kaggle.com/reminho/human-glioblastoma-dataset