methodological niche & short or long reads

[xjhzjucas]

Dear developer,
Thank you for the nice tool!
Here are my two questions regarding this software:

1.It is clear that the software operates in two steps: first, generating OTUs, and second, assigning taxonomic annotations to these OTUs against the GTDB database. My first question is whether the outcome of the first step can be considered conceptually equivalent to the methodology of the mOTU software. Furthermore, I have noticed that while mOTU was originally designed to be independent of reference genome constraints (e.g., GTDB), this software reintroduces GTDB in the second step, which makes me question its precise methodological niche. In this context, would it be valid to use only the results from the first step (i.e., the OTU table without GTDB annotation) for downstream analysis?

2.The software accepts both short and long reads as input. For my dataset, I have Cyclone long-read sequencing data (which does not fall under the recommended categories of Nanopore ≥ R10.4.1 or PacBio HiFi), complemented by 6 Gb of Illumina short reads for gap filling. Given this setup, I would appreciate your advice on what would be the most appropriate input for the software: the long reads, the short reads, or the assembled contigs?
Thanks!
Jinghong

[wwood]

Hi,

1.It is clear that the software operates in two steps: first, generating OTUs, and second, assigning taxonomic annotations to these OTUs against the GTDB database.

Right. There is also a 3rd step though - combining the results of the marker gene annotations into an overall community profile (the "condense" step in singlem.

My first question is whether the outcome of the first step can be considered conceptually equivalent to the methodology of the mOTU software.

Not really. The way the two pieces of software generate OTUs is quite different - in mOTUs the clustering happens on the reference database's marker gene sequences, while in SingleM it happens for each metagenome input. There are further differences e.g. how the clusters are defined too.

Furthermore, I have noticed that while mOTU was originally designed to be independent of reference genome constraints (e.g., GTDB), this software reintroduces GTDB in the second step, which makes me question its precise methodological niche.

Not sure I quite understand what you saying here sorry.

In this context, would it be valid to use only the results from the first step (i.e., the OTU table without GTDB annotation) for downstream analysis?

For SingleM, yes that is mostly true. Adding taxonomy annotations (step 2) doesn't change the OTUs, however, it might be worth doing so that you can exclude "off-target" sequences e.g. eukaryotic sequences or archaeal homologues to bacteria-specific genes.

2.The software accepts both short and long reads as input. For my dataset, I have Cyclone long-read sequencing data (which does not fall under the recommended categories of Nanopore ≥ R10.4.1 or PacBio HiFi), complemented by 6 Gb of Illumina short reads for gap filling. Given this setup, I would appreciate your advice on what would be the most appropriate input for the software: the long reads, the short reads, or the assembled contigs?

I've not worked with Cyclone data before (or DNBSEQ) so I'm not sure. SingleM works best with high quality sequence data (particularly if you want to trust the results of prokaryotic_fraction) but I would be curious if you had any observations. If they are sequences from the same sample, then they should be directly comparable, and so would be an interesting benchmark, right?

HTH, ben

[xjhzjucas]

Thank you very much for your prompt reply. I will try to perform some benchmarking analyses, and perhaps they may lead to some new findings.

I also have a few additional questions I would like to discuss with you. Please forgive me if some of them seem basic or a little unusual, as I am still not very familiar with this area.

1. How can we obtain the actual OTU sequences and use them to estimate abundance in a way analogous to TPM or RPKM? At present, the software output reports coverage, which seems different from the abundance measures typically used in conventional metagenomic analyses. This may make it difficult to perform standard downstream analyses such as beta diversity, since there does not appear to be a conventional abundance matrix. I was wondering whether it would be feasible to extract all sequences from the “sequence” column in the OTU table and then map the reads back to them in order to calculate abundance.

2. I would like to know whether the taxonomic information in the OTU results table can be used directly as evidence in a publication. I seem unable to find a clear bridge between the OTU table and the final profile output, so I do not know what ultimately happens to each OTU. The documentation states that the taxonomy in the OTU table is only preliminary, which makes me uncertain about how much confidence I should place in it. Is there a file or table that explicitly links each OTU to its final taxonomic assignment?

3. After reading both the paper and the software documentation, I am still unfortunately unclear about how “new lineages” should be defined. Can an OTU simply be regarded as a potential novel species-level lineage if it is not annotated to the species level in the OTU table, and similarly, based on the lowest resolved taxonomic rank, be interpreted as a potential novel genus-, family-, or higher-level lineage? I am not sure whether this would be an appropriate interpretation.

4. Are SingleM results directly comparable across different samples? For example, if I want to use OTUs to evaluate archaeal and bacterial diversity, can I directly compare the number of OTUs between two samples? Or would this be problematic because such comparisons may be affected by sequencing depth and related factors? Similarly, would it be appropriate to calculate alpha diversity based on OTU results together with abundance estimates?

[wwood]

Hi,

Good questions.

You can get the OTU table by running singlem pipe --otu-table (or for more detail use --archive-otu-table or include-extras. You can do alpha and beta diversity on these, but each marker should be treated separately (and use markers that target both archaea and bacteria, see singlem metapackage --describe.

If you want to use it in publications, I would suggest making sure your conclusions hold across many markers, not just one.

That doesn't provide a direct link betwee a coverage profile and the OTUs. There is a way to do that, but I think it might be too complicated as it is more of a debug thing.

I don't think just because you see a new sequence outside of the 96.7% assignment set that you would be able to conclude that it is a new species, because the data isn't quite clean enough (sometime there are erroneous reads, or strains with particularly divergent marker genes, and so on). You can get a sense of how many novel species you have by simply looking at the coverage of species-level lineages vs coverage of non-species-level lineages (see singlem summarise --output-taxonomic-level-coverage).

Working how much is novel above the genus level is not advised, because when not assigning to a species level, singlem uses a best-hit approach and so can over-specify taxonomy to a genus level when there is novelty.

HTH, ben

[susheelbhanu]

Hi all,

Sorry to be jumping on this thread, but similar to the query above, particularly in reference to this:

That doesn't provide a direct link betwee a coverage profile and the OTUs. There is a way to do that, but I think it might be too complicated as it is more of a debug thing.

I'd be interested in being able to track the contig name to the abundance and the taxonomy. Is this possible to do?

Thank you,
Susheel

---

Update:

Just figured out I could possibly do this with --output-extras. Is this correct?

[wwood]

Hi @susheelbhanu ,

You can use --output-extras but that is extra on top of the OTU table (e.g. it says what read/contig is in each OTU), but this is technically not the "profile".

The dev option I was talking about was actually condense --output-after-em-otu-table, but I guess this isn't what you are after.
ben

[susheelbhanu]

Thank you @wwood.

the first is what I was looking for, and managed to find it.