Understanding gemmeli's changes to a phylogenetic tree within phylo-CTF

[mestaki]

Hi @cameronmartino ,

This is a continuation of an issue first raised over towards the end of this gg2 thread.

The short of it is, I have a 150nt V4 feature (04195686f2b70585790ec75320de0d6f) of interest within the r220 GG2:
d__Bacteria; p__Proteobacteria; c__Gammaproteobacteria; o__Enterobacterales_A_737866; f__Enterobacteriaceae_A; g__Escherichia_710834; s__Escherichia fergusonii

I use gemelli to run phylo-CTF and load the community plot with the tree in empress. When I search for this feature in the tree, I see it is labelled differently as:
d__Bacteria; p__Proteobacteria; c__Gammaproteobacteria; o__Enterobacterales_A_737866; f__Succinivibrionaceae; g__; s__

The other 3 other branches you see below the selected node have the same "different" taxonomy too than their original taxonomy in gg2 tree.

So somehow the phylogeny label and gg2 taxonomy don't agree on this feature at the family level. I guess I'm trying to understand what phyto-CTF does to infer a different label. Is this an expected behavior?

[cameronmartino]

Hi @mestaki,

This might be a bug, I need to dig in a little deeper. The taxonomy is done through LCA via tax2tree. What version are you using? I think @wasade and I solved this here.

[mestaki]

Hey @cameronmartino,
I'm using version 0.0.12 of gemelli and 1.2.0 of empress.
Attaching the .qzv temporarily here in case helpful.
Sorry I missed the Q2 thread you linked before coming here, but, it does seem fairly similar to the issue here. I should mention there are other features mislabelled in the gemelli tree, I was just using this one as an example.

original commands:

!qiime gemelli phylogenetic-ctf-with-taxonomy \
    --i-table srs-11661-table-dada2-150-gg2-ADE_only.qza \
    --i-phylogeny gg2-filtered-tree.qza \
    --m-sample-metadata-file ../results/ctf-metadata_ADE_only.tsv \
    --m-taxonomy-file gg2-taxonomy-150.qza \
    --p-state-column emperor_time \
    --p-individual-id-column pig_id \
    --p-min-feature-frequency 10 \
    --output-dir diversity-results/gemelli/phylo-ctf-results-ADE_only

!qiime empress community-plot\
    --i-tree diversity-results/gemelli/phylo-ctf-results-ADE_only/counts_by_node_tree.qza \
    --i-feature-table diversity-results/gemelli/phylo-ctf-results-ADE_only/subject_table.qza \
    --i-pcoa diversity-results/gemelli/phylo-ctf-results-ADE_only/subject_biplot.qza \
    --m-sample-metadata-file ../results/ctf-subject-metadata-ADE-only.tsv \
    --m-feature-metadata-file diversity-results/gemelli/phylo-ctf-results-ADE_only/t2t_taxonomy.qza \
    --p-filter-missing-features\
    --p-number-of-features 50 \
    --o-visualization diversity-results/gemelli/phylo-ctf-results-ADE_only/empress.qzv

Thanks,

[wasade]

Tax2tree doesn't perform LCA and it is sensitive to the set of tips available.

If the tree has labels then decoration isn't necessary but modifications to the tree need to be done carefully to preserve the labels. If the tree does not have labels, it is important to decorate prior to any modification of the phylogeny (shear, prune, etc).

[cameronmartino]

@wasade We do currently shear the phylogeny as the first step of the import here. Actually tax2tree is the last thing called before return, after fastunifrac (here). Are you saying that could be the issue?

[wasade]

I hesitate to say yes or no, but the taxonomy generated is sensitive to the phylogeny