16S vs shotgun: what your microbiome test actually reads

When a company sequences your stool, it is making a choice you rarely see on the box — and that choice sets a hard ceiling on what your result can possibly mean. Almost every gut test runs on one of two technologies: 16S rRNA amplicon sequencing or shotgun metagenomics. They are not two brands of the same thing. They read different amounts of the underlying biology, and confusing them is the root of a lot of overinterpretation. Here is the difference, kept concrete.

16S: a barcode for “roughly who”

The 16S rRNA gene is present in all bacteria, with regions conserved enough to grab with universal primers and variable enough to tell groups apart. Sequence one or two of those variable regions and you get a cheap, fast census of which bacterial groups are present. The catch is resolution. As Johnson et al. (2019) documented, short-read 16S reliably reaches the genus level but only erratically distinguishes species — and almost never strains. It is also vulnerable to which region you target and to the fact that different bacteria carry different numbers of copies of the gene, which distorts the apparent proportions. Most important: 16S reads taxonomy, not function. It tells you who might be there, not what they can do.

Shotgun: read everything, including the verbs

Shotgun metagenomics skips the single barcode and sequences all the DNA in the sample — bacterial, viral, fungal. The canonical reference, Quince et al. (2017), lays out what that buys you: species- and often strain-level identification, and direct access to gene content — the metabolic pathways the community actually encodes, from short-chain fatty acid production to antibiotic-resistance genes. In a head-to-head, Durazzi et al. (2021) showed shotgun resolves lower-abundance taxa and reaches deeper than 16S on the same samples. The cost is real: more money, more DNA, and far more computation to make sense of the data. On the consumer side, the services that run shotgun through CLIA-certified, CAP-accredited labs — among them the personalized-probiotics company we profile here — sit at the higher-resolution end of this trade-off.

16S asks “who is roughly here?” Shotgun asks “who is here, and what genes do they carry?” A test can only answer the question its method was built to ask — no amount of dashboard design adds resolution the sequencing never captured.

The function on a 16S report is usually a guess

Plenty of 16S-based products still hand you a “metabolic” or “pathway” readout. How? Through prediction tools such as PICRUSt and its successor PICRUSt2 (Douglas et al., 2020), which infer likely gene content by matching your taxa to reference genomes. That inference is reasonable when your microbes closely resemble well-studied reference strains and degrades when they don’t. It estimates genomic potential, not measured or expressed function. A predicted pathway on a 16S report is a model’s best guess, not a thing the test directly observed — worth knowing before you act on a “you’re low in butyrate producers” verdict.

The caveat that outranks both: the method shapes the answer

Even with the right technology, the lab protocol can move the result more than your biology does. The Microbiome Quality Control project, Sinha et al. (2017), sent blinded specimens across 15 labs and 9 pipelines and found that DNA-extraction method, handling and bioinformatics — not just the sample — drove much of the variation. The consumer-facing version is starker: the NIST-led Servetas et al. (2026) study sent identical reference material to direct-to-consumer companies and found that differences between providers equalled or exceeded the differences between entirely different people’s guts. Same stool, different verdicts — covered in our piece on test reproducibility.

And you’re reading proportions, not amounts

One more structural caveat applies to both methods. Sequencing returns relative abundances that sum to 100%, not absolute counts — a point made forcefully by Gloor et al. (2017). If one microbe’s share goes up, others must mechanically go down, whether or not their actual numbers changed. A bug reported as “high” may simply be crowding the chart, not your colon. This is why a single percentage on a report card is one of the least reliable numbers in the whole field.

The honest bottom line

Neither method is “better” in the abstract; they answer different questions. If a test offers strain-level identification or genuine functional/metabolic gene content, it is almost certainly shotgun. If it’s priced cheaply and talks in genus-level bar charts — with any “function” inferred — it is almost certainly 16S, and you should read the metabolic claims as predictions. Either way, treat absolute-looking percentages and cross-brand comparisons with skepticism. For the bigger picture of what a result can and can’t tell you, see how microbiome testing works.