Inclusion/exclusion of levels of a factor
1
0
Entering edit mode
new_user6 • 0
@new_user6-20726
Last seen 4.5 years ago

I have the below data:

Samples        Genotype  treatment Time
D2_WT_mock1     WT       mock      D2
D2_WT_mock2     WT       mock      D2 
D2_WT_mock3     WT       mock      D2
D2_KO_mock1     KO       mock      D2
D2_KO_mock2     KO       mock      D2
D2_KO_mock3     KO       mock      D2
D2_WT_inf1      WT       inf       D2
D2_WT_inf2      WT       inf       D2
D2_WT_inf3      WT       inf       D2
D2_KO_inf1      KO       inf       D2
D2_KO_inf2      KO       inf       D2
D2_KO_inf3      KO       inf       D2

D6_WT_mock1     WT       mock      D6
D6_WT_mock2     WT       mock      D6
D6_WT_mock3     WT       mock      D6
D6_KO_mock1     KO       mock      D6
D6_KO_mock2     KO       mock      D6
D6_KO_mock3     KO       mock      D6
D2_WT_inf1      WT       inf       D6
D6_WT_inf2      WT       inf       D6
D6_WT_inf3      WT       inf       D6
D6_KO_inf1      KO       inf       D6
D6_KO_inf2      KO       inf       D6
D6_KO_inf3      KO       inf       D6

I have three different conditions, and would like to look at different comparisons for differential expressed genes. I combined the factors for easier interpretation. However when I look at the PCA plot (see attached), I do see a lot of variation within D6.

https://ibb.co/TvnWRS3

My goal is to see differences between genotypes based on treatment (using interaction terms), and main effects as well (for D2 and D6).

So my question is in this multi factor comparison, should I Run DEseq2 with all samples and remove the D6 ones by droplevels or run DEseq2 on D2 and D6 samples separately and make interpretations. Thanks.

If deseq2 is done together I could probably do comparisons between D2 and D6, if not that is alright if I just do these two separate.

deseq2 • 572 views
ADD COMMENT
1
Entering edit mode
@mikelove
Last seen 20 hours ago
United States

I would tend to run the models over the entire dataset. Some of the variation seen among D6 will be taken care of by the treatment and genotype coefficients.

ADD COMMENT
0
Entering edit mode

When I combine the factors to make interpretations easier, then I have several levels to compare via pairwise comparisons

 dds$combo
   [1] D2_WT_mock D2_WT_mock D2_WT_mock D2_KO_mock D2_KO_mock D2_KO_mock D2_WT_inf  D2_WT_inf  
   D2_WT_inf   D2_KO_inf  D2_KO_inf  D2_KO_inf  D6_WT_mock D6_WT_mock D6_WT_mock D6_KO_mock D6_KO_mock 
   D6_KO_mock D6_WT_inf  D6_WT_inf  D6_WT_inf  D6_KO_inf  D6_KO_inf  D6_KO_inf 

   Levels: D2_KO_inf D2_KO_mock D2_WT_inf D2_WT_mock D6_KO_inf D6_KO_mock D6_WT_inf D6_WT_mock

   res = results(dds, contrast=c("combo","D6_KO_inf","D6_WT_mock"))

Will it be biased if I do deseq of only D2 samples, and then drop levels of D6, and interpret following way

For D2,

     design(dds) <- ~ genotype + condition + genotype:condition
     dds <- DESeq(dds)

The effect of infection in wild type

     res_D2 = results(dds_D2, contrast=c("Genotype","KO","WT"))

The effect of infection in knockouts

      res <- results(dds, list( c("condition_inf_vs_mock","genotypeKO.conditioninf") ))

What is the difference between knock out and wild-type without infection?

      res = results(dds, contrast=c("genotype","KO","WT"))

this answers with treatment, what is the difference between knockout and wild-type?

      res = results(dds, list( c("genotype_KO_vs_WT","genotypeKO.conditioninf") ))

Similarly for D6, I will do the same list of comparisons?

ADD REPLY
1
Entering edit mode

Yes, these comparisons you list are fine, to perform these in D2 and D6 separately.

ADD REPLY

Login before adding your answer.

Traffic: 656 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