differential expression of miRNA

Question

[kamyasingh90]

Hi,

I am working on elucidating pathologically relevant miRNAs responsible for disease progression and chemoresistance in chronic lymphocytic leukemia. For differential expression of miRNAs I am using the edgeR package. I am new to statistics, edger and any help would be highly appreciated. The samples are classified as 1) Indolent which has 7 patient samples 2) Chemo-responsive, 9 patient samples 3) Non-responsive, 7 patient samples 4) Control, 2 samples. I want to check the differential expression of miRNAs as follows: (a) Control vs Indolent+Chemo-responsive+ Non-responsive (b) Indolent vs Chemo-responsive+ Non-responsive (c) Indolent vs Chemo-responsive (d) Indolent vs chemo-responsive. I went through the user guide of edgeR and in section 3.2.3 GLM approach by using contrast argument of glmLRT contrast(-1,1,0) will compare B to A sample. This works well when comparing two groups. How do I work when I have to compare A vs B+C and A vs B+C+D.

[kamyasingh90]

Thank you so much. This was very helpful for these conditions.

I was trying to look from clinical perspective for which I would like check the differentially expressed miRNAs in different sub-groups:

Indolent vs chemo-responsive, for which I used the contrast matrix as c(0,1,-1,0)

What contrast matrix should I use for :

1) Control(2 samples) vs indolent(7 samples),chemo-responsive(9 samples),non-responsive(7 samples). 2) Indolent vs chemo-responsive, non-responsive .

Answer 1

[Devon Ryan]

For A vs. B+C you're asking if A is different than the average of B and C, so c(1, -0.5, -0.5, 0) is your contrast matrix. Similarly, A vs the three others is c(1, -1/3, -1/3, -1/3). As an aside, you can swap the signs it you prefer to think of fold-changes as B+C+D vs A, as you did in your original example.

Answer 2

[Whoknows]

For miRNA analysis there are several good tools for analysis the best one in my idea is :

snRNAbench

This tools is the next version of miRanalyzer website, you could send your samples there and it performs analysis.

Another option is DESeq2 package, with this package your could define condition groups and also include time series information.

hope it helps.

[kamyasingh90]

Thank you so much for your help. I would try these options and hope it will help me. I still wanted to know if this can work with edge R. I guess i will face the same problem with DESeq2. I have already tried with deSeq. When comparing two groups I know the contrast matrix. But when I have to compare, for instance Control(2 samples) vs indolent(7 samples),chemo-responsive(9 samples),non-responsive(7 samples). 2) Indolent vs chemo-responsive, non-responsive . I don't know how to proceed. So any help regarding this would be very helpful.