miRNA reads alignment with bowtie2

[Ahaswer]

I've made some research before and found opinions that bowtie1 is not better than bowtie2 while considering miRNA data. So I'm convinced to use bowtie2 for this purpose, but still I have this problem.

[yueluo]

Don't use Bowtie1 unless you've checked it with simulated reads yourself. With perfect 20nt reads Bowtie1 has a 15% error rate compared to Bowtie2 (~2%)
http://genomespot.blogspot.com.au/20...-compared.html

Thanks for the input, so bowtie1 has the highest mapping rate but also significantly high incorrect rate. Sound like a terrible tradeoff.

[mziemann]

Don't use Bowtie1 unless you've checked it with simulated reads yourself. With perfect 20nt reads Bowtie1 has a 15% error rate compared to Bowtie2 (~2%)

http://genomespot.blogspot.com.au/2014/11/microrna-aligners-compared.html

[yueluo]

For miRNA data, I would suggest using bowtie instead of bowtie2, since the author claims bowtie to be more sensitive for reads <50bps.

Also, I would suggest you make a database from Rfam and ncbi/Ensembl's known non-coding RNA (mostly rRNA,tRNA,snoRNA, etc). Map reads against this first and you'll see how much contamination you have in the data.

[Ahaswer]

Hi everyone,

I'm working on miRNA data right now (which is very new field for me). I'm trying to filter out only those reads which belongs to miRBase. Here is my procedure.
Align all reads with bowtie2 (default settings) to the reference build index. Which gives me 92.23% overall alignment rate - looks good so far.
Building new index with miRBase hairpin data (with U/T conversion) and then bowtie2 alignment to this index. Last alignment provides only 5.17% overall alignment rate.
Also I've checked first alignment with genome browser using miRNA annotations which clearly shows that reads rarely align to mRNA so it should be fine.

Do you have any clue how to solve this problem? Is it common to have only ~5% coverage of miRNA sequences while data originate from miRNA sequencing?