[BioC] EdgeR: dispersion estimation

[Gordon K Smyth]

Dear Yanzhu,

I find it hard to believe that your data are all integers, because there 
is something seriously wrong, and we have only ever seen the problem you 
report when there are fractional counts.

How have you checked that they are all integers?  How were the counts 
created?  What do you see if you type:

   max(y$counts-floor(y$counts))

It also isn't clear that you have installed the current version of edgeR. 
Have you installed R 3.1.0?  Can you please give the sessionInfo() output? 
The current version of edgeR is 3.6.1, whereas you seem to be using 
version 3.2.4, which is two Bioconductor releases out of date.

Gordon

> Date: Tue, 29 Apr 2014 06:40:56 -0700 (PDT)
> From: "Yanzhu [guest]" <guest at bioconductor.org>
> To: bioconductor at r-project.org, mlinyzh at gmail.com
> Subject: [BioC] EdgeR: dispersion estimation
>
>
> Dear Gordon,
>
> Thank you so much for your reply and help. I greatly appreciate it.
>
> My data is exactly count data, which are integers. I got the same 
> results after I updated R/edgeR and rerun the code.
>
> I have some questions related to the dispersion estimation using edgeR 
> package. Briefly introduction of my multi-factor project (please refer 
> to my previous post for more details), I have three factors: L (16 
> levels), S (2 levels) and R(3 levels), so here I totally have 16 x 2 x 3 
> = 96 different conditions.
>
> 1. Due to some reasons, one of the condition has only one replicate, all 
> the other conditions have at least 5 replicates. In this situation, how 
> does edgeR estimate the common dispersion and tagwise dispersion?
>
> 2. I searched quite a lot about the examples using edgeR to estimate the 
> dispersion (including those examples shown in edgeR user guide), I found 
> that the common dispersion was not greater than 1 in most of cases, 
> however, I got 3.999943 for the common dispersion and 0.0624991 for all 
> of the tagwise dispersion. When the tagwise dispersions approach to same 
> value, shouldn't they be close to the common dispersion?
>
> 3. Use current version of edgeR, I tried different values for prior.df 
> (including the default) in the input of estimateTagwiseDisp function. 
> However, I always got the same results for tagwise dispersion estimates. 
> Are there any other input parameters in estimateTagwiseDisp function 
> that will affect the estimate results? And users can input the values 
> for these parameter according to their own projects?
>
> Please feel free to correct me if I make some mistakes here. I will also 
> greatly appreciate it if you can provide any other suggestions.
>
>
> Best regards,
> Yanzhu
>
>
>
>
>
>
> ###########################################################
>
> Dear Yanzhu,
>
> My guess is that some of your "count data" are not integers.  For example,
> are they perhaps expected counts from RSEM?  In the edgeR version that you
> are using, the GLM dispersion estimation functions do not work correctly
> for non-integer data.  (They weren't intended to.)
>
> Please update your copyies of R and edgeR to the latest versions.
> Bioconductor 2.14 was released a couple of weeks ago.  All edgeR functions
> now permit non-integer "counts".
>
> Also check that your data are counts and not RPKM or similar.  The counts
> should sum to the total sequence depth for each sample.
>
> Best wishes
> Gordon
>
>> Date: Wed, 23 Apr 2014 07:58:30 -0700 (PDT)
>> From: "Yanzhu [guest]" <guest at bioconductor.org>
>> To: bioconductor at r-project.org, mlinyzh at gmail.com
>> Subject: [BioC] EdgeR: dispersion estimation
>>
>>
>> Dear community,
>>
>> I use edgeR to do the data analysis of my RNA-seq project (as mentioned in my previous posts about multi-factor analysis of RNA-Seq project), I meet an issue with dispersion estimation:
>> I first used estimateGLMCommonDisp and then used estimateGLMTagwiseDisp to estimate the dispersion, however, I got 3.999943 for y$common.dispersion and 0.0624991 for all of the y$tagwise.dispersion (all of the y$tagwise.dispersion are the same). isn't it that all of the tagwise dispersion should NOT be the same?
>>
>> The fellowing is the code I used:
>> ##Read in count data
>> T<-data.frame(HTSeqRE)
>>
>> ##Factors:
>> Design<-data.frame(HTSeqCondRE[,2:4])
>> Rep<-as.factor(Design$Rep)
>> Line<-as.factor(Design$Line)
>> Sex<-as.factor(Design$Sex)
>> design<-model.matrix(~Line+Rep+Sex+Line:Rep+Line:Sex+Rep:Sex+Line:Sex:Rep)
>>
>> group<-paste(Design$Line,Design$Sex,Design$Rep,sep=".")
>> y<-DGEList(counts=T,group=group)
>>
>>
>> y<-calcNormFactors(y,method="TMM")
>>
>> y<-estimateGLMCommonDisp(y,design)
>> y<-estimateGLMTagwiseDisp(y,design)
>>
>> y$common.dispersion
>> [1] 3.999943
>>
>> y$tagwise.dispersion
>> [1] 0.0624991 0.0624991 0.0624991 0.0624991 0.0624991
>> 13474 more elements ...
>>
>>
>> Yanzhu
>>
>> -- output of sessionInfo():
>>
>>> sessionInfo()
>> R version 3.0.1 (2013-05-16)
>> Platform: x86_64-w64-mingw32/x64 (64-bit)
>>
>> locale:
>> [1] LC_COLLATE=English_United States.1252  LC_CTYPE=English_United States.1252    LC_MONETARY=English_United States.1252 LC_NUMERIC=C
>> [5] LC_TIME=English_United States.1252
>>
>> attached base packages:
>> [1] parallel  stats     graphics  grDevices utils     datasets  methods   base
>>
>> other attached packages:
>> [1] DESeq_1.12.1       lattice_0.20-27    locfit_1.5-9.1     Biobase_2.20.1     BiocGenerics_0.6.0 edgeR_3.2.4        limma_3.16.8
>>
>> loaded via a namespace (and not attached):
>> [1] annotate_1.38.0      AnnotationDbi_1.22.6 DBI_0.2-7            genefilter_1.42.0    geneplotter_1.38.0   grid_3.0.1           IRanges_1.18.4
>> [8] RColorBrewer_1.0-5   RSQLite_0.11.4       splines_3.0.1        stats4_3.0.1         survival_2.37-4      tools_3.0.1          XML_3.98-1.1
>> [15] xtable_1.7-3

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}