Artificial phenotypes

Data used to simulate artificial phenotypes from human gut metagenomes. <br> We used the same collection of human gut metagenome datasets as in Youngblut et al. (2020), with additional sample exclusion criteria and identical sequence processing. Samples with the following reported information were removed: i) samples from rectal swabs; ii) samples from individuals suffering from mumps, coeliac disease, gestational diabetes, cholera or with high relative abundances of Vibrio cholerae, infected by shiga toxin-producing Escherichia coli or cytomegalovirus; iii) samples with less than a million of sequence reads; iv) samples with missing age information. In total, metagenomes from 2147 samples from 19 studies and 23 countries were gathered. Microbial relative abundances were generated by Youngblut et al. (2020) using a custom database based on the Genome Taxonomy Database (GTDB), Release 89.0, created with the Struo pipeline. Only the relative abundances of families, genera and species with a prevalence above 25 % were included (p = 520 taxa). <br> ------------------------- X: data frame with relative abundances of taxa and number of reads for each sample <br> related_taxa_full: list of finer, coarser taxonomic level for each taxon <br> ruleExec_full: set of rules used to create the artificial phenotypes; rules were randomly drawn for each repetition.