Using vegan::betadisper for determing if I subset my design for DESeq2
1
0
Entering edit mode
@andrebolerbarros-16788
Last seen 26 days ago
Portugal

Hi everyone,

I'm currently analyzing a RNA-Seq data using DESeq2. Looking at the FAQ's in https://www.bioconductor.org/packages/devel/bioc/vignettes/DESeq2/inst/doc/DESeq2.html, concerning the subset of data, the authors recommend "users to run samples from all groups together, and then use the contrast argument of the results function to extract comparisons of interest after fitting the model using DESeq."

However, they show an example of when to subset: when we have one group that has a within-group variable higher than the remaining groups. Thus, I would like to know if we can use vegan::betadisper, a function to assess homogeneity of variances within-groups to determine if I should proceed or not with subsetting.

deseq2 • 876 views
1
Entering edit mode
@mikelove
Last seen 8 hours ago
United States

That's an interesting idea, and you could try it. I don't have any specific advice on whether that test is appropriate, but it looks like if you provided a distance matrix as we calculate in the workflow it would work.

I typically just look at PCA plots to determine if one group is very different in terms of its spread. And usually it's good to know this for all kinds of downstream applications and for biological interpretation.

0
Entering edit mode

Thanks for the feedback!

I got the idea from amplicon sequencing analysis, where they normally use PERMANOVA and, consequently, the betadisper functions.

This could be a good addition to help users to know if the groups have similar variances, besides EDA.

1
Entering edit mode

Hi,
Do you export the counts from the DESeq2 object to use in the PERMANOVA analysis? If yes, what kind of numbers do you use, normalised, vst, other?
Jon

0
Entering edit mode

Hey,

I'd say you would use the distance matrix based on the results from rlog or vst functions, the same you would use to make the PCA.

Cheers,

André

1
Entering edit mode

Of course, I see. Thanks!