DEXSeq sig-regulated exons only high in one sample

[Pedrissimo]

Hello,

I am comparing two biological conditions each condition has 2 biological replicates. The design matrix is:

samples = data.frame( + condition = c("high","high","low","low"), + replicate = c(1:2,1:2), + row.names = c("high1","high2","low1","low2"), + stringsAsFactors = TRUE, + check.names = FALSE + )

After creating my exon count set, I make a BA plot and color all the significant hits "padjust<0.05" in red. The vast majority of my sig-regulated exons are higher in my high sample with almost none in the low sample. A plot is attached, FPKM=read counts.

I have normalized my samples in a number of ways including the technique included with DEXSeq. I think my problem is that one pair replicates has a large variance and therefore the exons that are regulated in this direction are not detected?

I would like to test/demonstrate this by plotting the variance for EACH biological sample against the mean. Can someone show me how this might be accomplished.


TIA

[areyes]

Hi,

First of all... your plot looks really strange, why do you have negative FPKM/read count values?

What did you give as input to DEXSeq? You should give it raw count values per exon
can you send the commands that we use and your sessionInfo?

Alejandro

[Pedrissimo]

Hi Alejandro,

Sorry for the lack of clarity. from the table generated below, I take all padj values < 0.05 and graph their log2 foldchanges in blue, the non-significants are in grey. Here are the commands I issued:

annotationfile = file.path("hg19-chr.gtf")

samples = data.frame( + condition = c("high","high","low","low"), + replicate = c(1:2,1:2), + row.names = c("high1","high2","low1","low2"), + stringsAsFactors = TRUE, + check.names = FALSE + )

fullFilenames<- list.files("path to files",full.names=TRUE,pattern="counts.txt")

ecs<- read.HTSeqCounts(countfiles = fullFilenames,design = samples,flattenedfile = annotationfile)

ecs<- estimateSizeFactors(ecs)

ecs<- estimateDispersions(ecs)

ecs<- fitDispersionFunction(ecs)

ecs<- estimatelog2FoldChanges(ecs)

test<- testForDEU(ecs)

res1<- DEUresultTable(test)



> sessionInfo()
R version 2.14.1 (2011-12-22)
Platform: x86_64-unknown-linux-gnu (64-bit)

locale:
[1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
[3] LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8
[5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
[7] LC_PAPER=C LC_NAME=C
[9] LC_ADDRESS=C LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

attached base packages:
[1] stats graphics grDevices utils datasets methods base

other attached packages:
[1] multicore_0.1-7 DEXSeq_1.0.2 Biobase_2.14.0

loaded via a namespace (and not attached):
[1] hwriter_1.3 plyr_1.7.1 statmod_1.4.14 stringr_0.6 tools_2.14.1

[areyes]

Hi again,

The direction of the changes that you observe should not be affected by a higher variance in one of the biological replicates. What I still find strange are the negative count values in your MvsA plot. I would suggest to update to a newer version of DEXSeq, could you also send the output of

head(counts(ecs))

and an updated MvsA plot as the one in the vignette?

Alejandro