I'm working with the annotation of Illumina reads against mirBase and I have the reads counts. However this annotation gives me two alignments for the two types of miRNA (5p and 3p), however, when I want to compare my result with published result I found that in the article they don't use this discrimination and the count only the miRNA.
They sum the two result, 5p and 3p, or they choose only one, and why?
In the other hand, I read that first at all they make an alignment to the human genome and after that to the mirBase. Nevertheless, is it possible do the annotation directly from the trimmed fastq to the mirBase?
ADD COMMENT
• link
updated 3.1 years ago by
Ram
43k
•
written 9.5 years ago by
Kato
▴
60
1
Entering edit mode
In the article they might use the old annotation with or without a *. Conceptually you can distinguish between mir gene/ stem loop and mature miR based on the r and R.
Well you'll need to align the fastq file to something in any case. What you could do if you really don't care if it's the 5' or 3' mature product is align to the hairpin.fa file from mirBase. Regarding what they did in some paper, it's pretty much impossible to know unless they explicitly state things (or you can download their raw data and figure out how they got their results from them).
if this is miRNA analysis, the correct way to proceed is to differentiate between 5p and 3p. These two different small RNA that can come from the precursor have different function.
In that paper, maybe they used the old annotation, where the name is the main miRNA precursor that it could be 5' or 3'.
I recommend to use mirdeep2, or miraligner to get the counts of miRNAs, the second one is more directed to isomirs detection. Don't remember if mirdeep2 can use only mirbase, but miraligner uses only that. There are other onlines tools, like mirnabench that will do the job for you.
Just make sure ending with a counts for the mature miRNA if you want to have some functional information. If you don't, you can map directly with the precursor, but this are rare cases.
Good point, the change from the something* annotation to the current 5p/3p could easily be the cause of this.