Strangely Low Proportion Of Mapped Mirna Reads

Question

[Nick]

This is my first miRNA study. I'm focusing right now on differential expression of known miRNA genes. I am aligning the reads to the genome by running bowtie with the following parameters:

bowtie -n 0 -e 80 -l 18 -a -m 5 --best --strata

What I am concerned about is that only 10-30% of all reads map (percentage varies between the replicates). I checked few of the non-aligned reads and they seem to be mRNA. I presume the restrictive settings that I used prevented them from getting aligned to the genome.

I am doing just the bioinformatics analysis - shall I communicate to my wet lab colleague that this is a problem? Essentially my question is - shall we be worried about the low percentage of mappable miRNA in the samples? The quality of the data seems fine - judging the output of fastQC so it must be due to the way the biological material was captured and the library prepared. The study is on mice so I do not expect to stumble upon a lot of unknown miRNAs (which would have explained the low mappability). Let me know what you think.

[Rm]

try mirdeep2 and see how reads align

[Nick]

Thanks. I picked the above bowtie settings from the mirDeep2 paper. What I haven't done is look for novel miRNAs as I don't expect to see a lot of them in mice.

[Rm]

my best bet would be most of reads would be just the adapters? run fastx_clipper and see % of adapter only reads and too short reads after adapter removal?; also map to Rfam and see where you reads going other than miRNAs

Answer 1

[Jelena Aleksic]

Just to check - if this is from an sRNAseq experiment, what was the sequencing length used, and did you trim off both the 5' and the 3' end adapters? One potential source of mismatches would be the bits of adapter sequence.

However, if you're getting mRNA matches, sounds like that's not the issue - perhaps you could talk to the experimentalist about how they did the size selection?

[Nick]

Thanks for your answer. I'm waiting to hear from the lab regarding the adaptors. Another strange thing is that fastQC reports a lot of overrepresented sequences. I checked a few of them (e.g. CTCGCTGCGATCTATTGAAAGTCAGCCCTCGACACAAGGG) and they turned out to be mRNAs that are not unique to mice.