Fig. 3: Construction of protein core families from human gut microbiota.

a Mapping rates of metagenomic reads onto protein-coding genes with functional annotations. Violin plots show the distributions of annotation mapping rates across samples (n = 12,854 metagenomic sequencing runs). Over half of the genetic potential in a sample have functional annotations from MetaCyc or KEGG. Violin plots indicate median (white dot), the first and third quartile (black bar in the center), and the 1.5X interquartile ranges (black lines stretched from the bar). b Mapping rates of metagenomic reads onto protein families present in more than certain percentages of subjects in each metagenomic sequencing run (n = 12,854). All protein families (0% on the x-axis) accounted for 82.4% of the metagenomic reads. The core protein families were defined to be families found in >50% of subjects (50% on the x-axis), which accounted for 63.6% of the metagenomic reads. c Distribution of proteins across protein families. The core protein families in >50% of subjects represented 2.2% of all families, but included 62.7% of all proteins. The peripheral protein families in less than 10% of subjects represented 91.2% of all families, but included only 13% of all proteins. d Mapping rates of metagenomic reads onto the core protein families across the following developmental stages defined by the months of age: 3 to 10 (n = 4,645), 11 to 18 (n = 3,634), 19 to 26 (n = 2,252), 27 to 34 (n = 1,385), and ≥35 (n = 938). The mapping rates only decreased slightly as the subjects matured and their microbiomes diversified. e Principal component analysis (PCA) of the functional profiles of major orders over time. The functional profile of an order is the gene abundances of core families in this order in every KEGG category. f PCA of the taxonomic profiles of KEGG categories over time. The taxonomic profile of a KEGG category is the gene abundances of core families in this KEGG category in every order. Boxplots show the median (center), the first and third quartile (box), and 1.5X interquartile ranges (whiskers). Source data are provided as a Source Data file.