Extracting UMIs from NEXTFLEX Small RNA-Seq reads using UMI_tools

Question

[MH85]

Hello, everyone

I have miRNA single-end reads from the library prepared by NEXTFLEX Small RNA-Seq Kit. The read structure after trimming Illumina adaptors is:

| UMI 1 (4 nt) | miRNA | UMI 2 (4 nt) | Adaptor | Remaining sequence |

I want to use umi_tools extract to add UMI sequences to the read name. Can I achieve this by regex:

--bc-pattern=`(?P<umi_1>.{4}).+(?P<umi_2>.{4})(?P<discard_1>TGGAATTCTCGGGTGCCAAGG){s<=1}(?P<discard_2>.+)`

Thanks a lot in advance

Answer 1

[Kevin]

A few words of caution: just make sure you're using the v3 kit, as the v4 NEXTFLEX kit doesn't use randomized bases. Moreover, understand that the distribution of 8N bases for a given miRNA are non-random, since each miRNA exhibits preference regarding which subsequences it ligates with -- so you want to be careful using this sequence as a UMI.

[MH85]

Thank you very much! I wasn't aware that version 4 doesn't use randomized bases.

each miRNA exhibits a preference regarding which subsequences it ligates with

Isn't this a common issue with any ligation-based protocol? We currently use the QIAseq small RNA kit and have also observed ligation bias for certain miRNAs.

[Kevin]

Yes, there's bias introduced in every small RNA-seq method, since miRs have their own secondary structures (which may block miR ends) and there's also miRNA-adapter cofolding to consider. I just meant that the randomized bases themselves can't be assumed to have a random distribution, since each miRNA has its own preference of "random" bases. Qiagen adds their UMI's during reverse transcription, so that UMI should have more of a random distribution.

[i.sudbery]

To be fair, I've never seen a UMI dataset where the UMI usage looks random.

Answer 2

[i.sudbery]

Yes, that should work.