Using Generalized Linear Model in Microbial structure comparison

Question

[ohtang7]

Hi!

I am currently working to analyze alpha & beta diversity with microbiome data of 48 sample. (Animal stool samples)

I finished working of Beta diversity and got the data of significance. (Weighted UniFrac, Bray-curtis)

However, I received this message and need some help from you.

The message is as below (From reviewer)

I highly recommend the use of linear modeling (LM) or generalized linear modeling (GLM) which is commonly used in microbiome studies rather than a Wilcoxon rank-sum test. This will allow you to better control for things that may impact your results such as the age animals. It is also important to control for site as this is known to significantly affect the gut microbiome and could be included as a random variable in a generalized linear mixed model (GLMM).

I found that non-binomial regression or possion regression is the most commonly used glm for microbial analysis and trying to use the models in comapring Alpha and Beta diversity difference of the 16s rRNA data.

However, my question is that what package in R or python is recommended when figuring out microbial analysis. (Perhaps, many researchers in the metabarcoding field use some common statistical tools)

And, also I want to know that how can I modify (adding a dependent variable or etc. ) the data file to use in the GLM model. (What variable should be added to compare in Alpha and Beta diversity analysis, and how can I calculate them?)

my metadata file is as below

Any recommendation about using Package source, specific GLM model or the way to modify data - would be a big help for me to solve the problem.

Thank you

Answer 1

[Jean-Karim Heriche]

In R, linear models are available with the function lm() and generalized linear models with glm(), both from package stats. For mixed effects, you can use the lme4 package (functions lmer() and glmer()). There's also the function glmnet() from the glmnet package if you want LASSO or elastic net regularization.

Answer 2

[LauferVA]

Hello - in addition to the packages listed above (which I have used and they are good), there are also dedicated bioconductor packages written specifically for microbiota analysis.

A few that might work based on your post include: package DirichletMultinomial, package Microbiota, and package phyloseq.

Links to these 3 resources and dozens of other dedicated Bioconductor packages specifically for microbiota may be found here.

[ohtang7]

Thank you for your comment. However, your link does not work. Could you check it again?

[GenoMax]

Link should work now.

[LauferVA]

GenoMax - i assume you already took care of this for me since it appears to be working now - my apologies (maybe I missed part on the cut and paste, not sure) and TYVM for doing that. VAL

[ohtang7]

Thank you for your kind reply. I got many good information for my future study.