3
$\begingroup$

I have recently been tasked with analyzing a heterogeneous microbial sample composed of 16S data generated on a MinION sequencer, with the goal of assigning taxonomies and calculating respective abundance.

It is my understanding that traditional preprocessing methods such as Dada2, Unoise3, and Deblur would likely not be appropriate for this kind of data given the poorer read quality inherent with nanopore sequencing. To give an example, the average read quality of this current experiment is a Phred score of 15 and an average read length of 1500bp.

Furthermore, it would seem best to try and utilize a classifier that makes use of a full length 16S reference, rather than just the V3 region, to try and offset the higher error rate. But are there any classifiers that do this out of the box?

I am very interested to hear of any advice or input the community has on this. Thanks!

Improve this question
$\endgroup$
1
  • $\begingroup$ I'm doing similar work. Kraken2 with silva 138 db plus bracken works fine. Only constrain is that none reads were classified at species level. I'm wondering if you guys has obtained species level assignments? Cheers, Gaofeng $\endgroup$
    – Gani
    Jul 8, 2020 at 6:39

2 Answers 2

Reset to default
3
$\begingroup$

Use kraken2 with one of the 16S databases. Kraken2 has worked well for me with nanopore sequences. Pavian can be used to inspect the Kraken2 report files visually.

Using uncorrected called reads has worked for me. Any errors from nanopore sequencing should be inconsistent enough (i.e. not always lead towards the same wrong organism) that the noise doesn't matter.

Note: if you're using the MiniKraken database instead of a comprehensive 16S database, there might be some false positive species assignments depending on which kmers get pruned out.

Another possibility is centrifuge, but I'm not sure if that works for 16S amplicon sequencing; it's primarily for metagenomic analysis.

Improve this answer
$\endgroup$
2
  • $\begingroup$ @gringer conceptual doubt ..do we need perform denovo assembly on the nanopore data or we can simply use the fastq file for kraken2? $\endgroup$
    – kcm
    Jun 19, 2020 at 21:08
  • $\begingroup$ Using uncorrected called reads has worked for me. $\endgroup$
    – gringer ♦
    Jun 20, 2020 at 6:42
2
$\begingroup$

I would second the answer gringer gave recommending Kraken2. I would suggest using bracken to process the results from Kraken2 if you want to estimate species abundance (the raw Kraken2 results should not be used for this). I would recommend Pavian for visualising the results, it produces beautiful Sankey plots in a fairly pain-free manner. If you want to save yourself some work the 16S tool provided by Oxford Nanopore as part of epi2me produces good results for limited effort - although this is a bit of a proprietary black box.

Improve this answer
$\endgroup$
1
  • $\begingroup$ "lthough this is a bit of a proprietary black box." it takes a lot of time at one time only one barcode sample can be uploaded $\endgroup$
    – kcm
    Jun 19, 2020 at 21:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.