I’m working on human microbiota using the 16S analysis and I obtained a metabolic profile of my samples using picrust2 plugin of Qiime2.

I would kindly like to ask you, if you know some scripts or applications that can provide links between metabolic pathways, that are significantly changed, and that can clearly represent them.

Thank you very much!

Hi @Teonexone,

Do you mean that you want to see the differences of the metabolic pathways between samples?

I think you can try:

- https://bitbucket.org/biobakery/biobakery/wiki/lefse
- ANCOM function which already implemented in Qiime2, please see the moving picture tutorial in Qiime2

Hope that helps.

Thanks @Lei,

I’ve already use the methods you suggested but I have obtained too much results to have a clear overview on the metabolisms changing.

Hi @Teonexone,

Are all the results significant? May be you need to describe your questions more specific and in more details.

I am also not an expert of statistic. Let’s waiting for other people’s reply.

Hi @Teonexone,

I really like your question, as I’ve tried to do this too and couldn’t find any easy to use scripts or applications to answer these questions.

The last time I worked with picrust predictions, I calculated a simple log2-fold change between groups, took the most enriched pathways, and visualized them on a pathway map using iPATH 3.

https://pathways.embl.de/

Colin

Thank you @colinbrislawn

I’m trying to use iPATH3 as you suggested me and it could be what I was looking for.

Matt.

Maybe, something similar to categorize_by_function.py for Qiime1?

Hi @Teonexone @colinbrislawn

I have found myself in this exact situation after using PICRUSt2 to predict metagenomic function from my 16S data. I was considering using DESEQ2 to try and identify differentially abundant MetaCyc pathways between my two groups. However, calculating a log2-fold change between groups, and visualising the most enriched pathways is an interesting idea. Have either of you published results from this analysis since this thread? Or are there new alternatives I have missed?

Thanks

Alicia

Hello Alicia,

Yes! iPATH was used to make Figure 6) E on this paper, basically as suggested here.

I wish there was a cleaner way to do this, so I’m interested in suggestions too.

Colin

Thanks for the reply Colin! I got some interesting results using this approach so it is nice to see an example

Hello.

I wish to know if the log fold change was calculated against the actual reading of each sample from different groups? Or it was calculated in the mean of the readings?

Also, was the log2 fold change on the relative proportion( percentage)?

This seems to be a nice way of representing picrust data. Thus, I would really appreciate some insight on this.

Thanks