Correct DESeq2 interaction design and results with paired and nested data
1
1
Entering edit mode
@juliachariker-22693
Last seen 4.7 years ago

Hi, I am analyzing 24 samples that measure group (IN, IP) by time (naive, 2DPI, 10DPI, 42DPI). My design, sample table, and commands for extracting results are given below.

There are 12 animals, and each animal is measured at both levels of group. Three different animals are measured within each time period. After reading vignettes and posts on designs with paired and nested data, I created a new variable to reflect the nesting called animalID.nested and added it to the design below to explore main effects and interactions in the data.

This is the first time I have dealt with paired and nested data, and I wonder if I have dealt with it correctly. It seems correct given everything I have read, but I am concerned I may have missed something. I would appreciate any feedback.

DESIGN

~ group + dpi + dpi:animalId.nested + group:dpi

SAMPLE TABLE

animalID    animalID.nested group   dpi
1   1   IN  naive
1   1   IP  naive
2   2   IN  naive
2   2   IP  naive
3   3   IN  naive
3   3   IP  naive
4   1   IN  2DPI
4   1   IP  2DPI
5   2   IN  2DPI
5   2   IP  2DPI
6   3   IN  2DPI
6   3   IP  2DPI
7   1   IN  10DPI
7   1   IP  10DPI
8   2   IN  10DPI
8   2   IP  10DPI
9   3   IN  10DPI
9   3   IP  10DPI
10  1   IN  42DPI
10  1   IP  42DPI
11  2   IN  42DPI
11  2   IP  42DPI
12  3   IN  42DPI
12  3   IP  42DPI

EXTRACTING RESULTS FOR MAIN EFFECTS AND INTERACTION

IP VS IN AT EACH TIME POINT

results(dds, name="group_IP_vs_IN", test="Wald")      #IP vs IN for dpi naive (the main effect)
results(dds, contrast=list(c("group_IP_vs_IN","groupIP.dpi2DPI")), test="Wald")      #IP vs IN for dpi 2DPI
results(dds, contrast=list(c("group_IP_vs_IN","groupIP.dpi10DPI")), test="Wald")      #IP vs IN for dpi 10DPI
results(dds, contrast=list(c("group_IP_vs_IN","groupIP.dpi42DPI")), test="Wald")      #IP vs IN for dpi 42DPI

TEST BETWEEN TIME POINTS FOR EACH CONDITION.

results(dds, name="dpi_2DPI_vs_naive", test="Wald")      #IN_2DPIvsNaive
results(dds, name="dpi_10DPI_vs_naive", test="Wald")      #IN_10DPIvsNaive
results(dds, name="dpi_42DPI_vs_naive", test="Wald")      #IN_42DPIvsNaive
results(dds, contrast=list(c("dpi_2DPI_vs_naive", "groupIP.dpi2DPI")), test="Wald")      #IP_2DPIvsNaive
results(dds, contrast=list(c("dpi_10DPI_vs_naive", "groupIP.dpi10DPI")), test="Wald")      #IP_10DPIvsNaive
results(dds, contrast=list(c("dpi_42DPI_vs_naive", "groupIP.dpi42DPI")), test="Wald")      #IP_42DPIvsNaive

INTERACTIONS

interactionGroupEffectAcross2DPI<-results(dds, name="groupIP.dpi2DPI")
interactionGroupEffectAcross10DPI<-results(dds, name="groupIP.dpi10DPI")
interactionGroupEffectAcross42DPI<-results(dds, name="groupIP.dpi42DPI")
deseq2 Paired Nested • 954 views
ADD COMMENT
0
Entering edit mode
@mikelove
Last seen 1 day ago
United States

I don't have time to check over analysis code unfortunately. My recommendation would be either to cross check your code with a statistician or someone more familiar with these designs, or to make plots of individual genes and convince yourself that you're confident in what is being produced in each of the estimates.

ADD COMMENT
0
Entering edit mode

Hi Michael, My apologies. I would never expect you to do that, and perhaps I included way too much information in my question. I am mostly concerned that I handled the paired and nested data correctly. I have seen posts on dealing with paired data and posts on dealing with nested data, but I have both. I believe the following design handles both of those correctly given your previous answers and what I know about regression models. ...but I find that what you don't know can often get you into trouble! So I wanted to confirm that I had not missed anything in setting up this analysis. I do appreciate your time. Generally, most of us have more than we can handle in the time allotted. Julia

~ group + dpi + dpi:animalId.nested + group:dpi
ADD REPLY

Login before adding your answer.

Traffic: 525 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