time couse RNAseq experiment 2 treatments x 4 times
1
0
Entering edit mode
hakusen03 ▴ 10
@hakusen03-10599
Last seen 3 months ago

Hello, I have a time course RNAseq experiment (2 trt x 5 time points), and I want to find differences in gene expression levels over time for both treatments. I specified the experiment in terms of a factorial model:

design <- model.matrix(~trts * time, data=trtstr)


And, I got the following coefficients

"(Intercept)"      "trtsTRTB"        "time14d"          "time1d"          "time22d"          "time7d"          "trtsTRTB:time14d" "trtsTRTB:time1d" "trtsTRTB:time22d" "trtsTRTB:time7d"


I'm interested in the following contrasts:

my.contrasts <- makeContrasts(TRTAvsTRTB.0d = TRTA.0d-TRTB.0d,
TRTAvsTRTB.1d = (TRTA.1d-TRTA.0d)-(TRTB.1d-TRTB.0d),
TRTAvsTRTB.7d = (TRTA.7d-TRTA.0d)-(TRTB.7d-TRTB.0d),
TRTAvsTRTB.14d = (TRTA.14d-TRTA.0d)-(TRTB.14d-TRTB.0d),
TRTAvsTRTB.22d = (TRTA.22d-TRTA.0d)-(TRTB.22d-TRTB.0d), levels=designB)


Will qlf <- glmQLFTest(fit, coef=7:10) equivalent to these contrasts?

Also, will the model.matix account for sampling from the same individual at different times?

Thanks,

edgeR timecourse • 183 views
0
Entering edit mode
@gordon-smyth
Last seen 36 minutes ago
WEHI, Melbourne, Australia

The contrasts you are interested in are equivalent to coefficients 2, 8, 10, 7 and 11, respectively, in the factorial model.

If you don't mind me saying, wouldn't it be easier to fit a mean-effects model so you could form the contrasts explicitly (as suggested in the edgeR User's Guide)? Then you wouldn't have to ask what the coefficients mean because it would be obvious.

No, the model does not account for repeat measurements from the same individual.

0
Entering edit mode

Yes, it is easier to understand. I was making sure the contrasts were equivalent to coefficients used in the previous analysis. Thank you for the clarification! Do you have any suggestion on how to account for repeated measurements from the same individual in an RNAseq experiment?