This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Specifying model in DESeq - differentiation experiment

Hi all,

I am trying to figure out the best way to specify the following experiment in DESeq2.

I have a differentiation expression consisting of five conditions, the original set of starting cells (Ctrl Day 0) which are then differentiated by adding X or by a slightly different route by adding X and trt, with samples taken at two timepoints.

i.e.

  1. Ctrl Day 0
  2. Diff (X) Day 5
  3. Diff (X) Day 10
  4. Diff (X+ trt) Day 5
  5. Diff (X + trt) Day 10

And I'm interested in asking what changes over time in both Diff and Diff + trt, and which genes are significantly different between (Diff Day 5 versus Ctrl) and (Diff + trt Day 5 versus Ctrl) and for Day 10 aswell and comparing Diff Day y against Diff Day y + trt.

I'm a little unsure how to model this properly. I've tried various ways of doing it (treatment + timepoint, and coding treatment as two variables etc) and I'm not getting anywhere and am pretty stuck.

One option is to remove the Ctrl Day 0 sample completely but this feels like the wrong thing to do.

Any suggestions? Thanks in advance,

rna-seq deseq2 experimental design deseq
• 2,097 views • link
updated by Carlo Yague Carlo Yague • written by Nathan • Nathan

1 answer

What I would do :

          1    2    3    4    5
X :       NO   YES  YES  YES  YES
trt :     NO   NO   NO   YES  YES
time :    0    5    10   5    10

X is confounded with time, as (Day0 vs (Day5, Day10)) is the same as (noX vs yesX), so I would take only the time effect, the treatment effect and the interaction between both :

~time+trt+time:trt
• 0 views • link
written by Carlo Yague Carlo Yague

Thanks for the suggestion, but I'd already tried that approach and kept on coming up with the standard full rank model exception. I guess I'll more of a play with how best to separation conditions.

• 0 views • link
written by Nathan • Nathan

Log in to answer this question.

More posts like this