Identifying Metabotypes from Tensor Data

V. Skantze, M. Wallman, A. Sandberg, R. Landberg, M. Jirstrand, C. Brunius. 18 th International Conference of the Metabolomics Society in Valencia, Spain, June 19-23, 2022.


Metabolic response to diet shows large individual variation, which warrants tailored dietary recommendation i.e., personalized nutrition (PN). A step towards PN is to tailor diet to groups of individuals with similar metabolic phenotype, so called metabotypes (i.e., clusters of individuals with similar metabolism). Metabotyping of high-dimensional data is commonly performed in matrix form using matrix decompositions (e.g., PCA). However, data from e.g., crossover studies can be conveniently organized in multi-dimensional form (i.e., as tensor data) and methods for detecting metabotypes in such data are still lacking.

Photo credits: Nic McPhee