I’m intended to discover the correlation among metabolites from different bodysites, say, serum, vagina and gut. Should I use DIABLO, mixMC or multilevel in mixomics package?
I found the samples of DIABLO is on different omics of sample samples, while mixMC is designed for microbiome.
By the sounds of it, a multilevel approach isn’t going to be appropriate, unless you have multiple measurements of each individual at each body site.
While DIABLO is intended to take different omics types in each block, you could use different body sites as each of these metabolomics blocks. This would facilitate your assessment of correlation structure between each of these sites. However, for DIABLO you will need a categorical response variable. If you do not have access to this, explore
block.spls() rather than
If you are not measuring metabolites from all sites for every participant, you could just use
splsda() and have the body sites as the response variable (like this case study). If you have repeated measured, then look at this case study.
Regarding mixMC, when dealing with microbial data you need to apply the right pre-processing. Refer to this page of the website for more information on this.
Thanks Max. I do have categorical variable for subjects (healthy and diseased).
Are there any papers you know to do similar DIABLO on bodysites? Not limited to metabolome, other -ome like microbiome, immunome or proteome.
mixMC can be used to deal with metabolomic data?
Apologies, just realised I misunderstood your outline. mixMC should only be used with microbial data (contains OTUs or ASVs). For your metabolite data, mixMC processing won’t be appropriate.
I’m not aware of any specific papers regarding DIABLO on multiple bodysites unfortunately. I’ll keep an eye and an ear out though