[BioC] edgeR asimmetry in miRNA experiment

[Gordon K Smyth]

No, the unbalanced sample numbers will not cause edgeR to give asymmetric 
DE results.

Quite the opposite.  Since, miRNA results are expected to be globally 
assymmetric, I would expect edgeR to be under-stating rather than 
exaggerating the assymetry.

BTW, please upgrade to the current release of R, Bioconductor and edgeR.

Best wishes
Gordon



On Tue, 30 Apr 2013, Genomnia - Guffanti Alessandro wrote:

> Dear colleagues: I am analyzing the result of some cancer samples miRNA NGS
> experiments with edgeR, with the following setup:
>
> R version 2.15.3 (2013-03-01)
> Platform: x86_64-w64-mingw32/x64 (64-bit)
>
> locale:
> [1] LC_COLLATE=Italian_Italy.1252  LC_CTYPE=Italian_Italy.1252
> [3] LC_MONETARY=Italian_Italy.1252 LC_NUMERIC=C
> [5] LC_TIME=Italian_Italy.1252
>
> attached base packages:
> [1] stats     graphics  grDevices utils     datasets  methods   base
>
> other attached packages:
> [1] edgeR_3.0.8  limma_3.14.4
>
> I am following a rather standard workflow - please note that we have 8
> samples and only 1 control
>
>> targets <- read.delim("targets.txt", stringsAsFactors = FALSE)
>> d <- readDGE(targets, header=FALSE)
>> colnames(d) <-
> c("ARMS1","ARMS2","ARMS3","ARMS4","ERMS1","ERMS2","ERMS3","ERMS4","NMS")
>> dim(d)
> [1] 2038    9
>> keep <- rowSums(cpm(d)> 5) >= 3
>> dim(d)
> [1] 685   9
>> d$samples$lib.size <- colSums(d$counts)
>> d<-calcNormFactors(d)
>> d <- estimateCommonDisp(d, verbose=TRUE)
> Disp = 0.71417 , BCV = 0.8451
>> d <- estimateTagwiseDisp(d, trend="none",verbose=TRUE)
> Using interpolation to estimate tagwise dispersion.
>> de.com <- exactTest(d)
>> sum(de.com$table$PValue < 0.05)
> [1] 97
>> topValues <- topTags(de.com,n=97)
>> summary(decideTestsDGE(de.com,p.value=0.05))
> [,1]
> -1   44
> 0   641
> 1     0
>
> What we noticed is that there is a strong asimmetry in the corrected P
> values, in that only the downregulated miRNAs have a significant corrected P
> value - the upregulated miRNAs are less when examing the uncorreetd counts,
> basically we have alf of the CPM
>
> Questions:
>
> => is the unbalanced experimental design affecting the results ? this
> unbalance is coherent with the literature, in cancers the majority of miRNAs
> are downregulated
>
> => if yes, can I correct it or we should just take the results as they are
> and validate extensively if we want to explore also the upregulated miRNAs ?
>
> Thanks a lot in advance for any help,
>
> Alessandro & colleagues
> -----------------------------------------------------
> Alessandro Guffanti - Head, Bioinformatics
> Genomnia srl
> Via Nerviano, 31/B – 20020 Lainate (MI)
> Tel. +39-0293305.702 / Fax +39-0293305.777
> www.genomnia.com [http://www.genomnia.com/]
> alessandro.guffanti at genomnia.com [mailto:alessandro.guffanti at genomnia.com]

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:5}}