Yes, DIABLO is suitable for your data. The multivariate frameworks in mixOmics are not restricted to high-dimensional omics and can apply to any data set of continuous variables, including targeted panels of the size you describe.
One thing worth mentioning is that you may not need to select variables in the small panels (e.g. genes and fatty acid) and perhaps minimal selection in ht metabolites, or none.
Additional details were discussed in this previous post: