alpha diversity - variables


How can I view (generate alpha diversity boxplots) variables with multiple groups in alpha diversity?
Can I do that in qiime2?
To clarify: I have variable X consist of two main groups. I want to view those groups based on different age groups from the same dataset (without rerunning the dataset separately each time based on age groups). Is it possible?

In other words, I have two main groups to compare.The two main groups will be compared based on different age groups. (total of three age groups)
Do you select the sampling depth for the main groups 3 different times (different sampling depth for each age group) ?That way, we ensure that we retrieve many sequences without losing many samples.

Are there any other considerations that I should be aware of?
Thank you!

Hi @AYK,

I feel like you've got two issues here.

Without filtering your metadata (maybe generating a new file for each age group), there are a few ways you can solve this problem.

You could try creating a composite column in alpha diversity boxplots that would compare the groups. Then, look at the post-hoc test results for the paired samples.

In q2-longitudinal, there's an anova function that takes a linear regression formula. You could pass in an interaction term (age*exposure), although it can be hard to directly interpret that model. I think the interaction term there would tell you if being in a different age group changes the relationship between your exposure and alpha diversity compared to your default age group.

I don't know if you passed in a metadata file that had fewer samples than the alpha diversity if that would work. I haven't tried it, but if it does, that might be an option.

You could also go into another software. Alpha diversity will be behave nicely in pretty much any statistical software that makes you happy: R, excel, Stata, statsmodels/scipy, SAS, SPSS, etc. I know a lot of people like qiime2R for passing their files; you can also export the data.

Personally, I like to do one rarefaction and diversity calculation for my entire study. There's some stochasticisticity in rarefaction and I have at least one example where the rarefaction iteration changed our interpretation. So, I try to work with a single version of the rarefied table. (I'll also occasionally average across multiple rarefied tables.) Then, I do one alpha diversity and/or beta diversity calculation. The beta diversity, especially, can take a long time or require a lot of computing power, and so I only want to minimize the number of times I do those calculations. Beta diveristy is easy to filter in QIIME 2; (alpha diversity is a little bit harder, unfortunately)


1 Like

Thank you very much Justine! That is very helpful insight.
I understand it better now.

1 Like