Question: using limma and edgeR for differential expression of Drop-seq/10xGenomics scRNA-seq data
0
gravatar for Bogdan
11 weeks ago by
Bogdan580
Palo Alto, CA, USA
Bogdan580 wrote:

Dear Aaron, and all,

after going over the tutorial for detecting differential expression in scRNA-seq data

(https://bioconductor.org/packages/release/workflows/vignettes/simpleSingleCell/inst/doc/de.html)

thought that I could ask you please :

would the same pipeline (that includes edgeR or limma) apply to scRNA-seq data from Drop-seq/10X Genomics ?

given the fact that the number of UMI reads per gene is very low (eg 0, 1, 2, 3, ...10, ... ; very few genes having 10-20 counts). thanks a lot !

-- bogdan

ADD COMMENTlink modified 11 weeks ago by Aaron Lun25k • written 11 weeks ago by Bogdan580

You have to be more specific. What pipeline are you talking about? Are you referring to the marker gene detection? Or to the pseudo-bulk DE analysis?

ADD REPLYlink written 11 weeks ago by Aaron Lun25k

Hi Aaron, thank you very much for your help and for the questions.

i was referring to "part 6 : Using pseudo-bulk counts" :

https://bioconductor.org/packages/release/workflows/vignettes/simpleSingleCell/inst/doc/de.html

that describes the analysis on a SMART-seq2 dataset (416B cells).

It would be great to have again your insights and help. Thanks a lot !

ADD REPLYlink written 11 weeks ago by Bogdan580
Answer: using limma and edgeR for differential expression of Drop-seq/10xGenomics scRNA-
3
gravatar for Aaron Lun
11 weeks ago by
Aaron Lun25k
Cambridge, United Kingdom
Aaron Lun25k wrote:

If you add the counts together to create the pseudo-bulk samples, they won't be low any more. So there's no problem if you have enough cells to contribute to the sum in each sample.

Even if you don't have many cells, edgeR would still be able to deal with low counts, provided the dispersion is low (< 0.5). Very high dispersions cause some of the approximations in the QL framework to fail.

If you have low counts and high dispersions, then this is the worst case scenario. edgeR will not be happy, but the same could be said for other methods, so there's not much that can be done here.

ADD COMMENTlink modified 11 weeks ago • written 11 weeks ago by Aaron Lun25k

ok, wonderful, just wanted to double check and have a professional statistical advice ;)

thank you for writing a detailed documentation on assessing the differential expression.

ADD REPLYlink written 11 weeks ago by Bogdan580
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 171 users visited in the last hour