Microbiome data is typically high-dimensional, meaning that there are many more features than samples. Such data poses challenges for statistical analysis and when two or more high-dimensional datasets are integrated it gets even harder. At Clinical Microbiomics we approach multi-omics data integration through dimensional reduction, use of prior-knowledge and hypothesis testing.
Examples of integrative data analysis that we can offer can be found in the Pedersen et al. 2016 Nature article: Human gut microbes impact host serum metabolome and insulin sensitivity, where we integrated serum metabolome, shutgun microbiome and clinical data, and found that species that shared the potential for branched-chain amino acids (BCAA) biosynthesis were linked to serum levels of BCAA and insulin sensitivity measured by HOMA-IR. The framework is described in more detail in the Pedersen et al. 2018 Nature Protocols.
We stress that such multi-omics analysis should be thought of as hypothesis generating rather than a final result and as in the Nature article followed up by a verification experiment.