How should the order of microbiome data preprocessing be prioritized?

Question

[ssko]

Hello,

I have the microbiome count otu-table and I want to do an inferential/regression analysis. I want to present the results of my analysis at genus level. The literature says that the order of pre-processing can change depending on the purpose, but I want to be sure to get a consistent result. Is there any order of data pre-processing for this type of analysis: for example, is it recommended to do taxonomic aggregation at genus level first, then filter the genera by prevalence, then normalization/transformation, or is it recommended to do filtering first and then aggregation at genus level?

Thank you!

Answer 1

[hakimbazol]

Based on my experience, it depends on your dataset. Since you are talking about OTU, I assume your data based on amplicon sequencing, which usually only accurate up to genus level. Perhaps you can start your analysis from the common one, such as phyla, and then it goes to more specific level. If you have phenotype data, it also helps you quite much I think.

Hope it helps!

[ssko]

So, is it acceptable to transform filtered relative abundance data with clr tranformation, or is it recommended to use all unfiltered data for clr transformation?

[hakimbazol]

I think you should use all unfiltered data for clr transformation, and then use the transformation result to explore further about genotype-phenotype relationship for example. I found the publication that may become your reference.

https://microbiomejournal.biomedcentral.com/articles/10.1186/s40168-023-01747-z

Hope it helps!

[ssko]

I can't thank you enough for your great answers! hakimbazol