How to combine differentially expressed miRNAs to calculate molecular signature score in edgeR?

Question

[Björn]

I have 15 DE miRNAs between groups A and B. Is there a way to combine let's say DE of 5 miRNAs of interest to calculate the molecular signature ? The idea is to see if combined value of 5 miRNAs can differentiate Groups A and B ?

Zeng X, et al. (2012) Circulating miR-17, miR-20a, miR-29c, and miR-223 combined as non-invasive biomarkers in nasopharyngeal carcinoma. PLoS One 7(10):e46367.

The reference above is from qrt-pcr data but how to perform similar analysis in edgeR package ?

Answer 1

[Kevin Blighe]

Dear Björn,

Yes, build a gene signature with the miRNAs and use stepwise reression for the purposes of reducing the signature to a select few that are 'most' predictive of your groupings: A: Resources for gene signature creation

Please also see my R training niotes on my GitHub page, which also covers this topic: https://github.com/kevinblighe/Rtutorials

Your starting model would be something like:

glm(grouping ~ mir1 + mir2 + mir3 + ... + mir15, data=MyData)

Kevin