beta-diversity analysis on pathways/functions

[Marco Severgnini]

Dear all, I'm performing data analysis on a set of shotgun metagenome samples and, I was able to create tables in .biom format containing counts of reads supporting functions or pathways for each sample. Now I'm wondering if these tables could be analyzed by QIIME and, in particular, if I can apply a "standard" beta-diversity analysis on them, i.e.: calculate inter-sample distances (with non-phylogenetic metrics), perform PCoA, perform "adonis" test to find whether two experimental classes differ etc.

Can this be made for BOTH functions and pathways (thus, grouping functions to upper hierarchical levels) or from an microbial ecology point-of-view I'm trying to do something meaningless or (even worse) wrong?

Thank you in advance for any help.

All the best

Marco

[TonyWalters]

Hello Marco,

I think that would work-you could probably use any of the non-phylogenetic metrics with your table. The tree-based ones will work *if* you have a tree with tips that matches the 'OTUs' in your table.

The other caveat I can think of is if there are odd characters in the 'OTU' ids that might throw off some of the scripts.

You might just try a biom summarize-table
command with your table to see if it's reading your table for sample numbers/observations correctly.

-Tony

[Marco Severgnini]

Hi Tony, thank you very much for your quick reply. I actually did convert correctly the both the function (KEGG orthology records) and pathway ("ko" records) tables into .biom format with no error. However, I was wondering if performing beta-diversity analysis in the same way I normally do on taxonomies provides a meaningful result or not. In particularly, when I analyze pathway tables (where functions have been summed up to upper hierarchical levels).

I hope that my question is clearer now.

Thank you very much again.

Marco

[TonyWalters]

Hello Marco,

I'd say it's going to tell you if there are meaningful differences in the categorized categories/pathways. There may, however, be a lot of diversity between the samples in the unclassified group/pathways that can't be meaningfully distinguished since it's all collapsed as a group here.

To get around the classification limitation, you'd have to try something else, like the kmer approach in simka (https://gatb.inria.fr/software/simka/) for the metagenomic data.

I hope this helps,

Tony