Dada2: addSpecies for ITS data

[aliruizrodriguez]

Hello!
I am processing ITS data with DADA2 (primers ITS1F-ITS2). I am following recommendations from this paper (https://www.sciencedirect.com/science/article/pii/S1754504818302800?via%3Dihub) and for now, I am only analyzing sequencing results from mock communities. After assigning taxonomy against UNITE database I found out that for some species, I cannot resolve the taxonomy further than genus. This is especially striking with the Aspergillus genus, I am getting multiple ASV assigned to Aspergillus and when I blast the sequences (NCBI) they come back with the right taxonomy at the species level. I know the length of the amplicons sometimes does not allow for species identification, but I was wondering whether there is an option such as addSpecies (for 16S data against silva_species_assignment_v138.fa.gz database) for fungi. I couldn't find anything similar on the UNITE website.
Also, for assigning taxonomy, I am using General FASTA release UNITE database that includes singletons set as RefS (in dynamic files). Is that right or should I use the one that includes global and 97% singletons? I am a bit lost here.
Thanks upfront for your help!
Alicia.

[benjjneb]

Also, for assigning taxonomy, I am using General FASTA release UNITE database that includes singletons set as RefS (in dynamic files). Is that right or should I use the one that includes global and 97% singletons?

I am not aware of a rigorous comparison of the two for taxonomic assignment using ITS short-read sequences.

I was wondering whether there is an option such as addSpecies (for 16S data against silva_species_assignment_v138.fa.gz database) for fungi. I couldn't find anything similar on the UNITE website.

assignSpecies really was developed with short-read 16S sequencing in mind, and the specific way that it works (unambiguous exact matching) is appropriate for species-assignment to short-read 16S, but I just don't know how well it works in other marker-genes or Kingdoms. So again, I unfortunately can't give you very concrete guidance here. The reality is that most of my practical knowledge is based on bacteria. DADA2 and methods are equally applicable to fungi, but I just don't have that practical experience to give this kind of guidance.

Let me try tagging a couple people that I suspect know way more about your questions than I do. @gzahn @naupaka @dleopold

Thoughts from anyone else that has done more work with taxonomic assignment in fungi are welcome.

[gzahn]

My advice is to use the new UNITE+EUK database. Having those outgroups in
there is crucial since common ITS primers co-amplify lots of other stuff.
I've done comparisons between the databases and the RDP classifier is
perfectly happy to assign some metazoan reads as fungi if it doesn't have
those outgroups to train on. I haven't really seen any difference when
comparing the Singleton vs 97% versions of UNITE, so I don't have much to
say about that part. Just make sure you include those euk outgroups. It
makes a big difference!

I'd also echo the notion that assignSpecies isn't a practical approach for
the ITS regions. We've noted plenty of intraspecific variation.... Even
some fungi that have two different ITS1 variants within the same culture
(joys of dikaryotism!).

Geoff Zahn

On Tue, Mar 16, 2021, 15:11 Benjamin Callahan @.*>
wrote:

Also, for assigning taxonomy, I am using General FASTA release UNITE
database that includes singletons set as RefS (in dynamic files). Is that
right or should I use the one that includes global and 97% singletons?

I am not aware of a rigorous comparison of the two for taxonomic
assignment using ITS short-read sequences.

I was wondering whether there is an option such as addSpecies (for 16S
data against silva_species_assignment_v138.fa.gz database) for fungi. I
couldn't find anything similar on the UNITE website.

assignSpecies really was developed with short-read 16S sequencing in
mind, and the specific way that it works (unambiguous exact matching) is
appropriate for species-assignment to short-read 16S, but I just don't know
how well it works in other marker-genes or Kingdoms. So again, I
unfortunately can't give you very concrete guidance here. The reality is
that most of my practical knowledge is based on bacteria. DADA2 and methods
are equally applicable to fungi, but I just don't have that practical
experience to give this kind of guidance.

Let me try tagging a couple people that I suspect know way more about your
questions than I do. @gzahn https://github.com/gzahn @naupaka
https://github.com/naupaka @dleopold https://github.com/dleopold

Thoughts from anyone else that has done more work with taxonomic
assignment in fungi are welcome.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/benjjneb/dada2/issues/1301#issuecomment-800607487,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ADUAE4TBNO6WSHGL76KQ45TTD7CPPANCNFSM4ZIKFAMQ
.

[aliruizrodriguez]

Thanks a lot @gzahn for your advice. Is this one the database you recommended to use for assigning taxonomy "https://plutof.ut.ee/#/doi/10.15156/BIO/786370" , UNITE website, general FASTA releases, Taxon group, All eukaryotes?
Also, have you encountered the same results when analyzing ITS short reads with DADA2 than Pauvert et al 2019 ((https://www.sciencedirect.com/science/article/pii/S1754504818302800?via%3Dihub)? Overall, better results when analyzing only forward reads, no need to remove the primers, no chimeral removal?
Thanks again for your help, it is quite hard to find tutorials for ITS out there.

[naupaka]

@aliruizrodriguez I'd agree with @gzahn's suggestions. In addition to assigning taxa in DADA I also sometimes take the ASVs and do a command-line BLAST search against NCBI nt excluding environmental sequences and compare to results from UNITE. If the high-level patterns are reasonably similar, then I feel ok talking about them in a paper. But in general the rule of thumb with ITS is not to over-interpret anything because it's a challenging locus to work with for many reasons (including inter-individual and inter-specific variability and copy number variation as Goeff pointed out)... There is some taxonomic bias with ITS1F towards Ascomycetes, and UNITE has historically had better representation from mycorrhizal fungi, so a lot depends on the samples and what you're trying to figure out.

I have also found that generally reverse Illumina reads are lower quality and cause problems if you try to match them, and it can be helpful to only use the forward reads unless you have a really beautiful high quality sequencing run and you've chosen your primers carefully. I generally remove primers early in the pipeline and do the chimera removal, but chimeras are in general much less of a problem with ITS than with 16S.

[naupaka]

Oh, and another point -- if I understand correctly how the taxonomic assignment happens with the DADA functions, it will only give you a taxonomic assignment down to the level where it is confident. If you want species-level info, then I would find the appropriate reference sequences and work on building some high-quality alignments and phylogenies to be able to assess what fits where. The DADA functions are great for high level overviews of what you've got in big data sets, but if you care about species or even genus level placement for your specific research, then I'd take the sequences and spend time doing a more careful phylogenetic tree (ML/Bayesian) based analysis. Once you have this tree you can always feed it back in when you do your visualizations downstream in phyloseq.

[benjjneb]

Thank you @naupaka @gzahn Amazing.

[gzahn]

@naupaka out of curiosity, when you say "build a phylogeny" from ITS reads, do you mean using something like Jack Darcy's GarbageTree or the GhostTree approach? I have essentially no luck getting alignments from ITS. What's working for you?

[naupaka]

@gzahn no -- I mean pulling out ASVs from the same genus or _maybe_ same family, and then doing traditional MSA and phylogeny building. There's no rigorous way to align more broadly than that with short ITS reads that I'm aware of. You can always use BLAST to match a target sequence of interest against an exported ASV fasta to pull out close matches before making the tree.