Microrna Reads Alignment

Question

[GPR]

Hello, I have performed a comparative microRNA analysis using miRDeep2, getting about 15 million reads mapped (based on the intensities in the csv output files). I have repeated the analysis, this time using the Bowtie>DeSeq strategy but to my surprise the RSEQC tool, bam_stat.py indicates that only about 6,000 reads were aligned. The alignment discrepancy between miRDeep2 and Bowtie is puzzling to me because miRDeep2 uses Bowtie. I have tried Bowtie with and without clipping the adaptors, using the perl script from miRDeep2, getting the same discrepancy. Can somebody comment on this? Also, does anybody know how to convert the *.arf files from miRDeep2, to a *.sam or *.bam one? Thanks in advance. G.

[Lorena Pantano]

How are the name of the sequences you map w bowtie? are directly the reads? or the unique sequences? Normally reads are collapsed in unique sequences with this name: seq6x3456 where 3456 are the counts. If u map unique sequences with names u have to sum those numbers to get the total mapped. Also, are u mapping against the genome or precursors? It is very weird, also u can look for the command mirdeep use to map and see the parameters.

Answer 1

[Charles Warden]

Not sure about miRDeep.

If your institution has a license, novoalign has a miRNA setting that is specifically designed for small RNA-Seq. This is what I would generally recommend.

Nevertheless, I have some suggestions:

1) Just with novoalign won't work as well without the miRNA setting, you probably need to set non-default parameters to get optimal results from Bowtie.

2) From what I understand based upon the reference manual for miRDeep, it sounds like the .arf file is a results table (so, you probably can't directly convert that to a .sam or .bam file)

3) If you worried about the quality of the miRDeep results, there are other free tools for miRNA-Seq.

For example, here is a web basedtool that was recently published:

http://bioinformatics.oxfordjournals.org/content/30/3/434.short

[web link not working right now, but it looked interesting when it was working; sorry that I don't have a better suggestion]

You can check these results more closely match your miRDeep results or your Bowtie results.