miRNA read counts regarding DE

[vebaev]

Hi,
I'm using several software to annotate known miRNAs from deep-seq. Regarding differential expression, most of the software point only miRNAs present in all the samples and their fold change.

I was thinking what about for examples if some miRNA are present in my control replicas (2) and have read counts, and in the other 2 samples replicas this miRNA is not present at all?
Does it mean something or in DE analysis should be considered miRNAs presented on some level in all samples?

Thanks!

[Simon Anders]

What kind of software are you talking about? If you want to test for differential expression of miRNA between conditions, use the same tools as used for mRNA, i.e., DESeq, edgeR or BaySeq. Neither of these has any problem if counts are zero for some samples.

[vebaev]

I mean when I got that some miRNA are zero in one sample and a specific reads in other sample should I look them at all? Is it biologically meaningful when in one sample is zero?

[NicoBxl]

what is the number of reads in the other samples ?

[vebaev]

I have checked and the other reads are quite low, so I'm planning to discard these entries

[Simon Anders]

If you use a tool like DESeq, edgeR, BaySeq, you don't need to worry about this at all; the tools will judge themselves whether counts are too low to be significant; and for this, there it makes no difference whether the count is just very low or actually zero

[vebaev]

Yes, I'm exploring for example DESeq, but I have problems creating the tabular input format for my 4 samples (2 controls and 2 conditions).
I do not know that to use to create it, should be tabular with readcounts, but miRNAs corresponding reads should be indentified first....so Im confused...what should I pipe into DESeq?

Can you recommend me a tutorial or documentation, because I found all the things for DESeq with protein coding seqs where are more complex?

[NicoBxl]

the input of DESeq is a data.frame ( column : sample, row : miRNA )

ex:

         Sample1     Sample2         Sample3
miR-1      10              20              15
miR-2      100             30              109

[vebaev]

Thanks I will try that and post here my progress

[vebaev]

PS
should I worry how to list columns (order) sample1 2 3 , as they are 2 controls and 2 conditions?

[NicoBxl]

no

you've to read the DESeq doc : http://www-huber.embl.de/users/anders/DESeq/

[vebaev]

Thanks will do that now

[vebaev]

as a total number of reads for library it should be considered total number of mapped reads (sum of all mIR counts in column) or total number of reads from sequencing?

[NicoBxl]

as a total number of reads for library it should be considered total number of mapped reads (sum of all mIR counts in column) or total number of reads from sequencing?

same question

[vebaev]

From what I have read I have end up with this - the total number is probably the total bumber of mapped reads, please correct me if I'm wrong!