YAML for ☝️ for #C04NLA7QHH9 params: Archaea:

queryseqfile: 'archaeal_sequences.fna'
refphylogeny: '<https://figshare.scilifelab.se/ndownloader/files/45818916>'
refseqfile: '<https://figshare.scilifelab.se/ndownloader/files/45818919>'
taxonomy: '<https://figshare.scilifelab.se/ndownloader/files/45828279>'
model: 'GTR+F+I+G4'

Bacteria:

queryseqfile: 'bacterial_sequences.fna'
refphylogeny: '<https://figshare.scilifelab.se/ndownloader/files/45818913> bac120.newick'
refseqfile: '<https://figshare.scilifelab.se/ndownloader/files/45818925> bac120.alnfna'
taxonomy: '<https://figshare.scilifelab.se/ndownloader/files/45828282> bac120.taxonomy.tsv'
model: 'GTR+F+I+G4'

Note that there's also a way of writing a

samplesheet.csv

with the above information:

sample,alignmethod,queryseqfile,refseqfile,refphylogeny,model,taxonomy
bacteria-r08rs214,mafft,bacterial_sequences.fna,<https://figshare.scilifelab.se/ndownloader/files/45818925,https://figshare.scilifelab.se/ndownloader/files/45818913,GTR+F+I+G4,https://figshare.scilifelab.se/ndownloader/files/45828282>
archaea-r08rs214,mafft,archaeal_sequences.fna,<https://figshare.scilifelab.se/ndownloader/files/45818919,https://figshare.scilifelab.se/ndownloader/files/45818916,GTR+F+I+G4,https://figshare.scilifelab.se/ndownloader/files/45828279>

Use with

--input

.

Let’s go! In our system, it’s hard to identify the main pathogen using V4 amplicon sequencing as that region is quite conserved in Pectobacteriaceae. We have an ASV that we think represents pathogen reads, which the dada2 classifier puts tentatively in a related family, but BLASTing the sequence against the nt database returns only hits for our pathogen at the species level. So I’m hopeful this might give another line of evidence that the sequence we’re looking at represents the pathogen 🤞

I'm interested in what you find as I haven't had time to look at our data. A colleague mentioned yesterday however that when close relatives are in the database precision is not better with placement. But you'll get a tree.

I’ll let you know! I know that the pathogen genome is in GTDB, so we’ll see what happens

Note that the Sativa procedure removes 16S sequences that do not have a phylogenetic signal that corresponds to the genome's taxonomy. So, although your pathogen is in the GTDB doesn't necessarily mean it's in SBDI-GTDB. Pathogens are of course likely isolates, which means chances are high that it's 16S passes Sativa. In any case, you might want to check what's in the taxonomy file.

Assuming I’m looking at the right file - it gets a whole bunch of ASVs down to genus level. Looking at the per_query.tsv file, it looks like it can’t place them to the species (goodwinii) level as there’s another closely-related species (izadpanahii) it can’t distinguish from.

OK, so one step better than the DADA2 classifier, if one sees this from the positive side? Quite likely that V4 doesn't contain sufficient signal to confidently assign to species. How does

addSpecies()

in DADA2 fare (

species_exact

in

ASV_tax_species.tsv

)? If you don't get species from this method even if there are exact matches, it's because there's more than one exact match from different species.

dada2 with addSpecies() doesn’t even find the genus!

It's not supposed to AFAIK. It's just for species, and then only when it finds unambiguous exact matches. I saw above that

assignTaxonomy()

only finds family.

I’ve noticed a weakness with this approach compared to the DADA2 classifier (at least for my data) - as we work on plant samples, we have to deal with the fact that a lot of our sequences are organellar (mitochondria/plastid) - in some cases, up to 80% of a sample. DADA2 picks these out quite nicely with the Silva DB, but I guess organelle genomes aren’t part of the GTDB so it can’t place these sequences accurately.

The major drawback with phylogenetic placement is that it's far more work reconstructing a tree than to just collect sequences. If you look at this from another perspective it's also a possible quality check as sequences that carry erroneous taxonomy tags can be identified. This is what we do with GTDB to create SBDI-GTDB bytesting each 16s sequence against its phylogenetic position in the taxonomy tree. If it doesn't fit, it's thrown away. (This takes a lot of resources for GTDB, and I don't want to think of how much it would take for Silva.) Since the GTDB contains trees, we get them almost for free after the first vetting.

That’s what I’m doing (the filtering was already in place in my pipeline, in any case). Looking at it, the chloroplast genomes are pretty much singularly placed as Cyanobacteria (“d__Bacteria;p__Cyanobacteriota;c__Cyanobacteriia;o__Cyanobacteriales;f__Microcoleaceae”) - which is close to correct - but the mitochondrial sequences seem to get placed all over the place (see picture). We’re probably getting a diversity of mitochondrial sequences though from the various eukaryotic things in each sample (fungi, protists, etc.) whereas the chloroplasts should come solely from the host. But the fact that lots of them are not even being placed in Alphaproteobacteria gives me some pause - I wonder if DADA2 has misclassified any of these as mitochondria when they’re not.

I could imagine that mitochondrial V4 amplicons don't have enough signal for placement in a bacterial phylogeny. So, since you seem to have further evidence that they actually are mitochondria you're probably fine. You could look at the likelihoods for the placements though.

Currently running a quick DADA2 comparison against the Midori 16S database for confirmation. The thing that’s got me a little concerned is that for a lot of samples, about 80+% of sequences are mitochondrial - despite including host-specific mitochondrial and chloroplast PNA clamps to block amplification. So what I want to double check as much as anything is a) are all these approximately 1000 ASVs actually mitochondria, and b) if so, is there interesting taxonomic data there we could use (i.e. it’s not the host).

I'm afraid I'm no help with mitochondrial sequences. I don't think we get many with our samples and primers.