Downstream meta-analysis from IBD Plexus metagenomics data

Downstream analysis and statistics MaAsLin
1

Hi,

Request your inputs regarding a query on downstream analysis from fecal metagenomics data from the IBD Plexus repository.

In this instance, we have metagenomics data for cases (IBD) from the IBD Plexus repository, that we want to compare with the control dataset (non-IBD) from a published study.

We merged the relative abundance (gene families) values from both studies in R, and are interested in case-control analysis using Maaslin.

A huge thanks to Dr. Eric Franzosa @franzosa for his immensely valuable insights and guidance. Eric suggested that such an analysis will not be statistically sound, as it will be difficult to clearly distinguish whether observed differences are due to case-control or due to batch effects from the studies. Another option that Eric suggested was to analyze these data along with multiple studies, which may help mitigate some of the confounding effects.

Eric also suggested that it may be helpful to get further insights from the forum.

Request you to kindly share any further guidance regarding this analysis.

Thank you very much for your inputs and suggestions.

Sincerely,

Archana

2

Hi @jarchana09

Sorry for my slow reply on email but I’ll address your question here. I agree with Eric. You will not be able to decouple study effects from disease effects if the two are completely explained by each other. I think in this case you could try and include other studies to try and get at the effect (although you will need to make some assumptions that the bias in one study is independent of the others to some degree).

Sorry I couldn’t be of much more assistance.

Cheers,
Jacob

3

Hi Jacob @nearinj ,

Thank you so much for your insights, which are very helpful.

I guess it makes more sense in this instance to try including more studies, as Eric and you have suggested, with the caveats around bias from the studies.

Really appreciate Eric’s and your thoughts and guidance on my queries, and the inputs you both have shared will help guide our analyses.

Thanks a million.

Sincerely,

Archana

1 Like