2
$\begingroup$

I am trying to replicate some results of a scRNASeq experiment and, when I looked at the data provided by the author, I noticed that some of the counts in the expression matrix are represented as decimals.

Would this be because the authors uploaded a normalised count matrix or for other reasons? Also, please note that the authors used the 10X genomics cellranger pipeline for the processing of raw data.

I have provided a brief snippet of the data frame here:

structure(list(P1TLH_AAACCTGAGCAGCCTC_1 = c(0, 0, 0, 0, 0, 0, 
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0), P1TLH_AAACCTGTCCTCATTA_1 = c(0, 
0, 0.314759545319035, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 
0, 0.314759545319035, 0, 0), P1TLH_AAACCTGTCTAAGCCA_1 = c(0, 
0, 0, 0, 0, 0, 0, 0, 0, 0, 1.18024827050334, 0, 0, 0, 0, 0, 1.18024827050334, 
0, 0, 0), P1TLH_AAACGGGAGTAGGCCA_1 = c(0, 0, 0, 0, 0, 0, 0, 0, 
0, 0, 0, 0, 0, 0, 0, 0, 0.644905372424165, 0, 0, 0), P1TLH_AAACGGGGTTCGGGCT_1 = c(0, 
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.780522072779651, 
0, 0, 0), P1TLH_AAAGCAACAGTAAGAT_1 = c(0, 0, 0, 0, 0, 0, 0, 0, 
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)), row.names = c("RP11-34P13.7", 
"FO538757.2", "AP006222.2", "RP4-669L17.10", "RP5-857K21.4", 
"RP11-206L10.9", "LINC00115", "FAM41C", "RP11-54O7.1", "RP11-54O7.3", 
"SAMD11", "NOC2L", "KLHL17", "PLEKHN1", "RP11-54O7.17", "HES4", 
"ISG15", "AGRN", "C1orf159", "RP11-465B22.8"), class = "data.frame")

Also, I am quite confused of the structure they used - as I don't have much experience with single cell datasets. Just to give some further context:

The colnames of the "full" dataframe are coded from P1TLH_sequence_1 to P5TLH_sequence_1. There are 5 patients in the study.

I'm assuming that each of the colnames represents a cell for a particular patient while the rows are the genes.

If this assumption is correct, how would I do the conversion from colnames in the dataframe to the "cell names"? And lastly, why are there decimals in this expression matrix?

Thank you all for your time,

$\endgroup$
3
$\begingroup$

It looks like you have normalized counts. My guess is that they are log-normalized with some scale factor, probably 10000.

With the whole matrix you should be able to work out the scale factor for each cell. See https://www.nxn.se/valent/2018/10/25/unscaling-scaled-counts-in-scrna-seq-data

They typical format for scRNA-seq data is to have cells in the columns and genes in the rows. So here your rownames are the gene names and colnames are the cell names.

$\endgroup$
  • $\begingroup$ Thanks for the response ! Do you have any idea of a DB of some sorts that is used to convert the colnames from sequence format to cellID format? $\endgroup$ – h3ab74 Feb 27 at 20:02
  • $\begingroup$ What's the cellID format? $\endgroup$ – TimStuart Feb 27 at 20:52
  • $\begingroup$ It is not a particular format called cellID per ce. I just mean conversion from the sequences to a naming system that would be more suitable for downstream analysis / clustering. $\endgroup$ – h3ab74 Feb 27 at 21:49
  • $\begingroup$ Ok in that case any unique identifier is fine and I would leave the cell names as they are. The only thing you might like to do is extract the patient ID information that appears to be encoded in the cell name. If you use Seurat, this will be done for you when creating the Seurat object. $\endgroup$ – TimStuart Feb 28 at 0:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.