Selecting the correct analysis

Hello. I am at the beginning of analysing a clinical dataset with 4 distinct groups (n=25 per group) of which 3 have an underlying pathology with different treatment outcomes, and the last one is the control group. I have shotgun metagenomics data, non-targeted GCMS metabolomic data, ~ 20 blood and metabolic measures and multiple clinical & physical measures. I want to investigate whether any of the variables can predict treatment response or whether any of the biological variables are associated with each other.
I have been reading the mixomics book and it’s very helpful, but I am still not entirely sure which is the method best fitted for my study. I think perhaps a multiblock PLS will be the most suitable, but would really appreciate your expert opinion.
I hope this makes sense. Thanks for your time!

Multiblock PLS will be useful to learn relationships between your various blocks of data. However what you’re really looking for is Multiblock sPLS-DA - also referred to as DIABLO. The DA means discriminant analysis and essentially refers to classification. Read all about it: methodology and case study.

However, as done in that case study, using PLS (or multiblock PLS) prior to DIABLO is a good way to explore your data and gain intuition about it. This is highly recommended