Recommended QIIME 2 workflow for ONT full-length 16S data and integration of EMU results for alpha/beta diversity

General Discussion
1

Hello. I am currently working with full-length 16S rRNA amplicon sequencing data generated with Oxford Nanopore Technologies (ONT), and I would really appreciate some guidance regarding the most appropriate workflow to process these data in QIIME 2.

The experiment was performed using: Full-length 16S rRNA gene amplicon, ONT Native Barcoding Kit for 96 samples, MinION flow cell FLO-MIN114, Basecalling performed with the most recent Dorado version available at the time of processing, Adapter/barcode trimming and read cleanup performed with Porechop. Although the expected full-length 16S amplicon is approximately 1.5 kb, after preprocessing I still have reads ranging approximately from 300 bp to 1650 bp.

My first question is therefore related to the appropriate QIIME 2 workflow for these already preprocessed ONT reads. Should I simply import the cleaned FASTQ files into QIIME 2 as sequence data, or would you recommend an additional filtering/denoising step before taxonomic classification?

In particular, I would appreciate recommendations regarding:

  1. Whether there is currently a recommended denoising/error-correction strategy within QIIME 2 for ONT full-length 16S data.

  2. Which taxonomic classification approach is currently considered most appropriate for this type of dataset. For example, would you recommend a reference alignment/classification approach, VSEARCH, BLAST, a q2-feature-classifier method, or another strategy specifically suitable for Nanopore full-length 16S reads?

In parallel, I have analyzed the same samples using EMU, and I would like to compare the taxonomic profiles and diversity results obtained with EMU and QIIME 2. EMU generates, among other outputs, a TSV file containing the read assignment distributions. According to the EMU output description, each row corresponds to an input read, the columns correspond to taxonomic IDs, and each value represents the likelihood that the read originated from that taxon. The values in each row sum to 1.

My goal is to use QIIME 2 to calculate alpha and beta diversity metrics, particularly phylogeny-dependent metrics such as weighted UniFrac and unweighted UniFrac. However, I am not sure about the best way to generate the corresponding QIIME 2 feature table and representative sequences (rep-seqs) from the EMU results. One possibility I have considered is to use the EMU read-assignment probabilities to determine the most likely taxonomic assignment for each read, separate the original reads according to those assignments, and then map them against the reference database using Minimap2 to recover the sequence/taxonomy relationship. From there, I could potentially construct a feature table and representative-sequence file for QIIME 2.

However, this seems rather indirect, so I am wondering whether there is a simpler or more appropriate method.

Thank you very much for your help.

1 Like
2

Hello!

I would not duplicate EMU analyses in QIIME 2 since EMU handles Nanopore read taxonomy annotation pretty well. But I would rerun EMU with the GTDB database instead of the native database - it contains a lot of outdated taxonomy names.

Also, if I remember correctly, EMU by default reports only relative abundances for each detected taxon, but you can adjust it to produce estimated raw counts. I would use those estimated raw counts at the species level to import them into QIIME 2 to proceed with alpha/beta (non-phylogenetic) analyses and DA tests.

Hope that helps.

Best,

3 Likes
3

Thank you very much for your help! I’ll follow your recommendations.