Visualization for Time Course "LRT" test in DESeq2
1
0
Entering edit mode
rbenel ▴ 20
@rbenel-13642
Last seen 1 hour ago
Israel

Hi, I am analyzing a time-course experiment and I am using the "LRT" test as I am interested in changes at any time point throughout the experiment. I would like to visualize this perhaps by using a Volcano plot to "show" the changes between the first day of the experiment and the last; however, I know that I can't use the fold changes associated with the "LRT" test as they are not directly associated with the actual hypothesis test, so is it okay to use the Wald test for this visualization purpose and color only significant genes that were found with padj < 0.05% in the results(dds_lrt)? Is there a better way to visualize the "LRT" results?

dds_lrt <- DESeq(dds_lrt, test="LRT", reduced = ~ 1)

res_lrt <- results(dds_lrt)

deseq2 time-course • 771 views
0
Entering edit mode
@mikelove
Last seen 1 day ago
United States

Have you seen the plot that we make in the workflow section on time course?

https://bioconductor.org/packages/3.7/workflows/vignettes/rnaseqGene/inst/doc/rnaseqGene.html#time-course-experiments

0
Entering edit mode

Oh I see, you want to see all genes at once. We don’t have a visualization scheme for this.

0
Entering edit mode

Yes, I would like to look at all genes (a more global picture).. What do you think of my suggestion to color only significant genes according to "LRT"?

0
Entering edit mode

I don't see what's going to go on the x-axis of the Volcano plot, if you are performing an LRT and there are multiple coefficients that are removed from the full design. There's not a single effect size to put on the x-axis.

0
Entering edit mode

You are correct , maybe I'm better off using a heatmap like in the tutorial...

Thanks!

0
Entering edit mode

I have a follow up question... as I am still looking for a visualization method for all deferentially genes at once. What if I compute all pairwise comparisons using test = "Wald", and I represent this as a heatmap of differential expressed genes in at least one time point. If there added value in representing this? I want to avoid representing 11 (the number of time points in my data) MA or volcano plots..

1
Entering edit mode

I guess the downstream plots are up to you. A heatmap of effects sounds reasonable.