I have a dataset with 3 conditions (control and 2 treatments), 2 replicates each. Initially I carried out a normal analysis in DESeq - each treatment vs the control, but later was told it needed to be a paired analysis.

Thus, just to be consistent I decided to use DESeq itself. However, I am unsure how to extract my comparisons of interest.

------

Code:

design

condition pair1 Cont 1

2 Cont 3

3 Trt1 1

4 Trt1 3

5 Trt2 1

6 Trt2 3

cds2 <- newCountDataSet(gct, design)

cds2 <- estimateSizeFactors(cds2)

sizeFactors(cds2)

Cont_1 Cont_3 Trt1_1 Trt1_3 Trt2_1 Trt2_3cds2 <- estimateSizeFactors(cds2)

sizeFactors(cds2)

0.9373964 1.1328686 1.0097990 1.0596419 0.8562104 1.0645546

.cds2 <- estimateDispersions(cds2,"pooled-CR",modelFormula=count ~ pair + condition)

fit1 = fitNbinomGLMs(cds2, count ~ pair + condition)

fit0 = fitNbinomGLMs(cds2, count ~ pair)

str(fit1)

'data.frame': 4765 obs. of 6 variables:fit0 = fitNbinomGLMs(cds2, count ~ pair)

str(fit1)

$ (Intercept): num 10.19 9.99 6.89 6.48 4.46 ...

$ pair3 : num -0.3018 -0.3222 0.068 0.0468 0.1504 ...

$ conditionTrt1: num 0.17 0.142 -0.495 0.125 0.319 ...

$ conditionTrt2: num 0.1882 0.00112 -0.2704 -0.09412 0.10651 ...

$ deviance : num 0.0742 0.1569 0.9072 2.0612 0.3702 ...

$ converged : logi TRUE TRUE TRUE TRUE TRUE TRUE ...

- attr(*, "df.residual")= num 2

head(fit1)

(Intercept) pair3 conditionTrt1 conditionTrt2 deviance convergedgene0 10.188336 -0.30184662 0.1698471 0.188197344 0.0741656 TRUE

gene1 9.992372 -0.32221027 0.1415406 0.001119304 0.1568869 TRUE

gene10 6.893693 0.06795915 -0.4954365 -0.270402107 0.9072335 TRUE

gene100 6.477847 0.04684995 0.1250743 -0.094116036 2.0611685 TRUE

gene1002 4.463446 0.15038277 0.3186187 0.106512332 0.3701619 TRUE

gene1003 4.090079 -0.05247162 -0.3560189 -0.054765642 1.9631768 TRUE

pvalsGLM = nbinomGLMTest( fit1, fit0 )

padjGLM = p.adjust(pvalsGLM, method="BH" )

---------padjGLM = p.adjust(pvalsGLM, method="BH" )

Based on the above how do I extract:

1) Trt1 vs Cont? - is this the 3rd column of fit1, depicted as log2 FoldChange?

2) Trt2 vs Cont? - is this the 4th column of fit1, depicted as log2 FoldChange?

3) What would the p-values for each comparison be? As pvalsGLM and padjGLM gives only one set, and I don't think it will be the same for both comparisons.

Many Thanks,

Natasha