multifactor desing with interaction LRT or Wald test?
Entering edit mode
jnaviapelaez ▴ 10
Last seen 18 months ago

Hi I am analyzing a RNAseq experiment with 2 factors,

I have been reading vignettes and other questions but I am still confused on what to do

donor(genotype) with 2 levels "WT" and "imKO", condition with 3 levels "naive" "cisplatin" and "cispAIBP",

My design = ~donor +condition + donor:condition

and my ref levels, dds$condition <- relevel( dds$condition, ref="naive" ) dds$donor <- relevel(dds$donor, ref="WT"),

ddsCollapsed <- DESeq(ddsCollapsed)

  1. I want to know if there is an effect of conditions in each genotype (main effect)?
res_cisp <- results(ddsCollapsed, test= "Wald", contrast=c("condition", "cisplatin", "naive"), alpha=.05)


res_AIBP <- results(ddsCollapsed, test="Wald", contrast=c("condition", "cispAIBP", "cisplatin"), alpha=.05)


However, If I do this how can I extract or compare the DEGs in only one genotype, for instance, WT?, I don't want to know the effect of each condition regardless of genotype but most the effect inside each genotype level. will the idea below work? I understand rescisp and resAIBP are comparisons of my first level (WT) and then for other genotype:

res_cisp_KO <- results(ddscollapsed, test="Wald", contrast=list( c("condition_cisplatin_vs_naive","donor_imKO.condition_cisplatin") ))
res_AIBP_KO <- results(ddscollapsed, test="Wald", contrast=list( c("condition_cispAIBP_vs_cisplatin","donor_imKO.condition_cispAIBP") ))

however, my results are not computing "condition_cispAIBP_vs_cisplatin" should i relevel my condition factor? in which case, should I create a second dds "ddsCollapsed2" with the new levels using cisplatin as reference and run DESeq to extract resAIBPKO ?

> resultsNames(ddsCollapsed) 
[1] "Intercept"                              "donor_imKO_vs_WT"                      
[3] ""   ""
[5] ""   ""

I think using a grouping variable might be an easier way to do it. but then the other question that I have is to know if there is a difference in condition effect across genotype, and that is why I included the interaction in the design. , it seems like my DESeq function results without arguments res <- results(ddsCollapsed) is doing it

but again, I don't know how to extract "" "" DEGs., based on ?results I think the following might work. but I don't really understand why. i don't think I understand resultsName output either Bytheway

res_cisp_ko_vs_wt <- results(ddsCollapsed, name="donor_imKO.condition_cisplatin")
res_AIBP_ko_vs_wt <- results(ddsCollapsed, name="donor_imKO.condition_cispAIBP")

resAIBPkovswt in this case is comparing AIBP effect but taking the first level "naive" as reference right? should I relevel?

and I also want to compare "donorimKO _vsWT" but at condition "naive" which is the reference level. is this the comparison in resultsNames(ddsCollapsed) ? If not I don't know how to do it. and I think the grouping variable will help here too.

can I use 2 different designs so I can do the grouping for my comparison and then including the interaction for the genotype effect on the main condition effect? OR I should do an LRT instead and in that case which should be my full and reduced design?

I appreciate any help,

deseq2 LRT wald • 153 views
Entering edit mode
Last seen 8 hours ago
Republic of Ireland

I answered a similar question on Biostars, recently:

There is also information in the DESeq2 vignette and via ?DESeq2::results (scroll to the bottom when the manual entry page comes up to see the examples, 1 to 3), which should help you to arrive at a decision. There are also many questions already answered on this topic on both Biostars and here on Bioconductor.

Trust that it helps.



Login before adding your answer.

Traffic: 263 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6