Untrimmed reads in small-RNA seq datasets

Question

[russell.stewart.j]

What are the typical percentages of raw reads that remain untrimmed in Illumina HiSeq 2500 small-RNA sequencing datasets? I cannot find any papers that state both their raw read and processed read counts to get an idea. I'm working with sequencing sets from testis tissue, and leaving 16-20% of my reads untrimmed. Library prep was done with NEBNext® Small RNA Library Prep Set for Illumina®, making the adapter -AGATCGGA...3'

Thanks!

[Brian Bushnell]

I'm not sure if there is a 'normal', considering how much variability there can be in library creation, and other factors. Also, I've never seen NEBNext data; I've only received emails from people who work with it. So I'll ask a few questions -

1) How are you doing your trimming?
2) What's the length distribution afterward?
3) What do you expect the length distribution to be?
4) Do you expect a large percentage of your reads to be off-target?  For example, PhiX or other spike-ins.
5) Are the reads paired?

[russell.stewart.j]

1) How are you doing your trimming?

• Using a custom script which trims everything after AGATCGGA. Also tried trimmomatic. Both result in 18% left over 51 nt untrimmed reads

2) What's the length distribution afterward?

• Length distribution is as I would expect, with strong miRNA and piRNA peaks.

3) What do you expect the length distribution to be?

• Peaks at ~21 and ~29 nt, corresponding to the above.

4) Do you expect a large percentage of your reads to be off-target? For example, PhiX or other spike-ins.

• In retrospect, I wish we had included spike-ins. Unfortunately we didn't use any.

5) Are the reads paired?

• I didn't realize that small-RNA datasets could have paired reads given their length, but no, not paired.

Thanks for your consideration, if you have any ideas please let me know.