Seurat/DESeq2 - single cell RNA-seq differential expression
1
0
Entering edit mode
Clara • 0
@deut2016-16915
Last seen 5 weeks ago
Germany

Hi all,

I am approaching the analysis of single-cell RNA-seq data.

I have seen that Seurat package offers the option in FindMarkers (or also with the function DESeq2DETest) to use DESeq2 to analyze differential expression in two group of cells.

Assuming I have group A containing n_A cells and group_B containing n_B cells, is the result of the analysis identical to running DESeq2 on raw counts of each gene in n_A versus n_B samples? And is there a way to speed up the analysis when n_A and n_B are in the order if a few thousands cells?

In addition, I have a 'technical' question:

when I have a count table in the form of a data.frame (for example read with read.table from a text file), is it necessary to force it to matrix such as cts <- as.matrix(cts) before providing it as input to DESeqDataSetFromMatrix? It seems to be I can just provide the count table as a data frame.

Thanks,

Claire

seurat deseq2 • 4.5k views
ADD COMMENT
0
Entering edit mode
@mikelove
Last seen 1 day ago
United States

 

As to speed, DESeq2 will be a lot slower than say a linear model when you have thousands of replicates. The software wasn’t optimized for this use case. According to a recent paper from Soneson and Robinson, you can use Wilcoxon effectively here for DE.

If you want to compare hundreds of cells with DESeq2, please use the zinbwave integration pipeline:

http://bioconductor.org/packages/devel/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#recommendations-for-single-cell-analysis

Finally, if you provide a data frame, DESeq2 I believe just does the conversion to matrix internally.

ADD COMMENT
0
Entering edit mode

Thank you! I had seen in this paper by Soneson&Robinson https://www.nature.com/articles/nmeth.4612 that DESeq2 was employed, however the number of cells were much lower. In addition, input used was transcripts per million, while I think Seurat uses raw counts. In any case, I will look at the recommendations you point out, which I had missed.

Regarding the internal matrix conversion, then this was maybe already present in DESeq2 versions from years ago as I had seen using a data frame in some old code as well.

ADD REPLY

Login before adding your answer.

Traffic: 518 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6