Extract reads that didn't contribute to miRNAs following miRDeep2 analysis

Question

[jammydodger123456]

As I understand it, the mapper.pl aspect of miRDeep2 will aim to map reads to the genome of interest with the help of bowtie, but not based on any structural signitures of a miRNA. These (5,776,111) reads are later filtered based on miRNA characteristics, giving those that look like miRNAs (3,265,557), and those that don't (2,510,554)

I would like to extract these 2,510,554 reads for further analysis (ie attempting to map against an index of other smRNA species). Is there any way to do this?

miRDeep2.pl relevant readout below

Converting input files
building bowtie index
mapping mature sequences against index
mapping read sequences against index
Mapping statistics

#desc   total   mapped  unmapped        %mapped %unmapped
total: 5776111  3265557 2510554 56.536  43.464
Dog: 5776111    3265557 2510554 56.536  43.464

Answer 1

[GenoMax]

You could manually run bowtie and collect unmapped reads as noted in answers here --> Separate unaligned reads in fastq format using bowtie

[jammydodger123456]

Yes but surely at the point of determining reads according to miRNA characteristics, this is past any bowtie mapping?

[GenoMax]

If you have a file that shows which reads were mapped then you could use the read headers to filter them out of the original file. filterbyname.sh from BBMap suite can help with that.