Adjust for covariates

[blacruz17]

Hi,

I'm working with microbial co-occurrence networks from human gut microbiome data. I've built them using NetCoMi with SPIEC-EASI as a network construction method. I want to compare network features across patient groups that have different phenotypes but also differ in age and gender. Is there a recommended way to perform a covariate-adjusted SpiecEasi analysis to control for these effects? If not, is there any network construction algorithm on NetCoMi that allows this?

Thanks!

[stefpeschel]

Hi,

thanks for your question and please excuse the very late reply! I will still answer your question since it might be relevant for other users as well.

NetCoMi currently does not provide a direct option to adjust for covariates such as age or gender during network construction. This also applies when using SPIEC-EASI through netConstruct().

None of the network construction methods currently implemented in NetCoMi includes a general covariate-adjustment step. If covariate adjustment is required for the network estimation itself, this would therefore have to be handled outside NetCoMi or by using a network inference method specifically designed to incorporate covariates.

I would be hesitant to recommend a general preprocessing approach such as regressing out the covariates, since whether this is appropriate depends on the data and the scientific question and is not equivalent in general to fitting a covariate-adjusted network model.

NetCoMi can, however, still be used for subsequent network analysis if an appropriate association matrix is estimated externally.

Best,
Steffi