Table 1.

Properties of metagenomic features and detection thresholds

Feature description	Short name	Sequencing basis	Units	Confident detection threshold	Relaxed detection threshold	Confident nondetection threshold	Body sites	Paired samples per body site
Operational taxonomic units	OTUs	16S rRNA gene	Relative abundance	>1e−3	>1e−4	<1e−5	18	25–105
Microbial species	Species	Whole metagenome shotgun	Relative abundance	>1e−3	>1e−4	<1e−5	6	14–50
Species-specific marker genes	Markers	Whole metagenome shotgun	RPKM	>5	>0.5	<0.05	6	14–50
Kilobase windows from microbial reference genomes	kbwindows	Whole metagenome shotgun	RPKM	>5	>0.5	<0.05	6	9–45

In analyses of per-feature stability, a feature was considered detected if its abundance exceeded the confident detection threshold; a feature was considered acquired if it initially fell below the confident nondetection threshold and then later exceeded the confident detection threshold. When defining a metagenomic code, features with abundance between the confident detection and confident nondetection thresholds were considered ambiguous. When reevaluating a code at a later time point, the relaxed feature detection thresholds were used to add robustness to temporal variation.