DESeq2 with three factor experimental design
1
0
Entering edit mode
microPhD • 0
@microphd-24044
Last seen 6 months ago

I'm running DESeq2 on RNAseq data and my experimental design has three factors: Carbon_Source, Time_Point, and Amoxicillin. I'm having trouble extracting the comparisons that I need.

Here is a portion of my sample metadata:

dds = DESeqDataSetFromMatrix(control_table, Design, ~Carbon_Source + Time_Point + Amoxicillin + Carbon_Source:Amoxicillin)

dds$Carbon_Source = factor(dds$Carbon_Source, levels=c("glucose", "dextrin"))
dds$Time_Point = factor(dds$Time_Point, levels=c("T0", "T1", "T2"))
dds$Amoxicillin = factor(dds$Amoxicillin, levels=c("none", "amox"))

DDS <- DESeq(dds)

resultsNames(DDS)
[1] "Intercept"                            "Carbon_Source_dextrin_vs_glucose"
[3] "Time_Point_T1_vs_T0"                  "Time_Point_T2_vs_T0"
[5] "Amoxicillin_amox_vs_none"             "Carbon_Sourcedextrin.Amoxicillinamox"


These are the comparisons that I think I have correct:

#Is the effect of amoxicillin different depending on glucose vs dextrin? (across timepoints)
res1 = results(DDS, name = "Carbon_Sourcedextrin.Amoxicillinamox")

#The effect of amoxicillin on the glucose condition:
res2 = results(DDS, name = "Amoxicillin_amox_vs_none")

#The effect of amoxicillin on the dextrin condition:
res3 = results(DDS, list( c("Amoxicillin_amox_vs_none", "Carbon_Sourcedextrin.Amoxicillinamox") ))

#Difference between glucose and dextrin conditions without amoxicillin:
res4 = results(DDS, name = "Carbon_Source_dextrin_vs_glucose")

#Difference between glucose T1 without amoxicillin and glucose T0:
res5 = results(DDS, name = "Time_Point_T1_vs_T0")

#Difference between glucose T2 without amoxicillin and glucose T0:
res6 = results(DDS, name = "Time_Point_T2_vs_T0")


I would also like to have the following comparisons, but I'm not sure how this can be done:

1. Difference between glucose T1 with amoxicillin and glucose T0
2. Difference between glucose T2 with amoxicillin and glucose T0
3. Difference between dextrin T1 with amoxicillin and dextrin T0
4. Difference between dextrin T2 with amoxicillin and dextrin T0
5. Difference between glucose and dextrin conditions with amoxicillin

I'm not sure that these comparisons are even possible since T0 only had one treatment type (no amoxicillin). This is also why I can't do an interaction term with time_point because the model matrix would not be full rank.

One idea I had was to run DESeq2 again with the same design, except flip the baseline levels so that "dextrin" is the baseline for Carbon_Source and "amox" is the baseline for Amoxicillin.

DESeq2 RNASeq • 171 views
0
Entering edit mode
@mikelove
Last seen just now
United States

If you have questions about the interpretations of the results, or how to properly choose a design, I'll recommend that you collaborate with a local statistician who can guide the analysis. Unfortunately I have to reserve my time on the support site to software related questions.