Hi, Sean:
 
Thx for the reminder. I usually included those info, just too many info for a single email. And turns out to be the key this time. Thx and best
 
Ming
 
Date: Sun, 9 Mar 2014 16:29:01 -0400
Subject: Re: [BioC] Questions about multi-factor contrast setting in DESeq2
From: sdavis2@mail.nih.gov
To: yi02@hotmail.com
CC: michaelisaiahlove@gmail.com; bioconductor@r-project.org




On Sun, Mar 9, 2014 at 1:11 PM, Ming Yi <yi02@hotmail.com> wrote:

Hi, Mike:



Thx for your advice again and I did try what you suggested as below:



> dds <- DESeqDataSetFromMatrix(countData = countD,colData = colD,design = ~Type + RasType + Type:RasType);

> dds <- DESeq(dds);

estimating size factors

estimating dispersions

gene-wise dispersion estimates

mean-dispersion relationship

final dispersion estimates

fitting generalized linear model

105 rows did not converge in beta, labelled in mcols(object)$betaConv. Use larger maxit argument with nbinomWaldTest

> show(resultsNames(dds));

[1] "Intercept"                   "Type_Tumor_vs_Normal"

[3] "RasType_RasP_vs_RasOnly"     "RasType_RasP1Hit_vs_RasOnly"

[5] "RasType_RasPNot_vs_RasOnly"  "TypeTumor.RasTypeRasP"

[7] "TypeTumor.RasTypeRasP1Hit"   "TypeTumor.RasTypeRasPNot"



certainly we can use

> res_RasP_vs_RasOnly <- results(dds,"RasType_RasP_vs_RasOnly")

> res_RasP_vs_RasOnly <- results(dds,name="RasType_RasP_vs_RasOnly")

> res_Tumor_vs_Normal<-results(dds,"Type_Tumor_vs_Normal")

> res_Tumor_vs_Normal<-results(dds,name="Type_Tumor_vs_Normal")



but I can not do the following with contrast as suggested in section 3.2:

> res_Tumor_vs_Normal<-results(dds,contrast=c("Type","Tumor","Normal"))

Error in results(dds, contrast = c("Type", "Tumor", "Normal")) :

  unused argument (contrast = c("Type", "Tumor", "Normal"))

>  res_RasP_vs_RasOnly <- results(dds,contrast=c("RasType","RasP","RasOnly"))

Error in results(dds, contrast = c("RasType", "RasP", "RasOnly")) :

  unused argument (contrast = c("RasType", "RasP", "RasOnly"))



also I can not get contrast like Tumor.RasP_Tumor.RasPNot:

> res_Tumor.RasP_Tumor.RasPNot<-results(dds,contrast=c(0,0,0,0,0,1,0,-1))

Error in results(dds, contrast = c(0, 0, 0, 0, 0, 1, 0, -1)) :

  unused argument (contrast = c(0, 0, 0, 0, 0, 1, 0, -1))



it seems the interaction terms in the design (design = ~Type + RasType + Type:RasType) changed the behavior of results()?



Hi, Ming.
Be sure to include the output of sessionInfo() when you report errors.  The information from that command tells us what version of R and OS you are using and what packages and package versions you have loaded.  

Sean

 		 	   		  
	[[alternative HTML version deleted]]