Weighted and Unweighted distance matrix have only zeros

[rgloreto]

Hi folks -

I am running into a issue with one of the beta analyses I have done.

I selected 8 OUTs from my data to run a separate analyses with them. After running the core diversity metrics, the distance matrix for both weighted and unweighted analyses returned are composed only by zeros. My collaborator ran the same data on Q1 and it worked well.
In addition, when I open the .qzv file, it returns a p-value of 0.001, even though the table only has zeros (see print screen).

I have redone this analysis quite few times and I can't figure out if it is the nature of my data (no diversity at all) or if it is a data format/code issue (since the collaborator was successful on running the same data on Q1).
FYI, before rung this data, I have ran the full data set and it indicated me this taxonomic group (8 stains of same bacteria) was the driving difference on my data.

Any insights?

Many thanks!

Best,
R.

[Nicholas_Bokulich]

Hi @rgloreto,
Could you please provide:

1. the precise commands that you are running (and ideally the filtering commands and any other relevant steps that lead up to the core metrics command)
2. run feature-table summarize on the feature table that you are using as an input to core-metrics
3. If you are able to share your files, sending those to us now may assist in solving this issue. You can send those in a direct message if you do not want to post on this thread for all the world to see. :eyes

@rgloreto

I can’t figure out if it is the nature of my data (no diversity at all)

Possible, but unlikely if your collaborator is doing the exact equivalent analysis in qiime1 with different results (though "exact" is highly unlikely as you may be getting different features).

@rgloreto

FYI, before rung this data, I have ran the full data set and it indicated me this taxonomic group (8 stains of same bacteria) was the driving difference on my data.

I suppose this is largely a matter of opinion and you probably have good reasons for this, but the goal of this analysis (filtering data and running beta diversity tests on a tiny subset) seems a little roundabout — there is probably a better way. If these 8 taxa are differentially abundant between groups, it would be much more direct to, e.g., generate boxplots or other plots showing the distribution of relative abundances for these taxa in each group. PCoA plots/permanova will simply indicate something that you already know — that these taxa distinguish your groups — and provide no/little additional information. PCoA plots/permanova are most useful when you cannot narrow down your features to a short list of differentially abundant taxa. With as few as 8 OTUs, it would be much more informative to examine their abundances directly. But again, that's just my opinion.

I hope that helps!