Cross domain interactions

[TeeratatK]

It is a short question. Is it practical for cross domain interactions as SpiecEasi package?

[stefpeschel]

You can generate the association matrix using SpiecEasi and use NetCoMi for network analysis and plotting as described here.

[TeeratatK]

Thank you for replying to my question. I'm working with 2 phyloseq objects that contain ITS data and Metabolite data, and I merged them into one object named "fun.bac.merge.renamed". I would like to ask you if I pass association matrix creation using SpiecEasi and use "measure = measure = "spieceasi" " instead. What is the different result that will be compared to the above tutorial?

net_genus <- netConstruct(fun.bac.merge.renamed, measure = "spieceasi", #Association measures measurePar = list(nlambda=10, lambda.min.ratio = 1e-2), normMethod = "clr", zeroMethod = "none", sparsMethod = "none", thresh = 0.3, #Defult verbose = 2 )

[stefpeschel]

Passing the association matrix to netConstruct() and setting measure to "spieceasi" won't work, because that would mean you are estimating associations from an association matrix, which doesn't make sense.

You have two options:

• Estimate the association matrix in advance with SpiecEasi and pass the matrix to netConstruct() as explained in the tutorial. Here you have to set dataType to "condDependence" (or just "correlation"; there's currently no difference).
• Pass your phyloseq object to netConstruct() and set measure to "spieceasi" so that associations are estimated within netConstruct(). In this case, the dataType must be "counts".

The latter won't work if you want to do a cross-domain analysis, so I recommend following my tutorial, which covers exactly your use-case.