Overview

This tutorial walks through a complete experiment using metabolomics data from the Franzosa et al. 2019  inflammatory bowel disease cohort. The same dataset is used in the microbiome tutorial, but here the metabolite abundance profiles replace the taxonomic profiles as the primary input modality.

The tutorial is labelled generic because metabolomics serves as a concrete example of working with a modality other than microbiota taxonomic profiles. The three configuration changes it demonstrates are passing the data through primary_data_file, setting features_are_rows=False, and leaving taxonomic_configs empty. These changes apply to any modality where the microbial taxonomic resolution step is not used. The metabolomics dataset was chosen because it comes from the same cohort as the microbiome tutorial, making the two a direct methodological comparison: same samples, same prediction task, different feature type.

Everything else, including the cross-validation strategy, the transformation search, the model search, and the output structure, is identical to the microbiome tutorial.

Steps