Ask a microbiome question and the first fork in the road is the sequencing strategy: amplicon (usually 16S rRNA) or shotgun metagenomics. The choice shapes everything that follows — what you can resolve, whether you can say anything about function, how much it costs, and which analysis pipeline you run. This guide compares the two honestly, so you pick the method that answers your question rather than the one that happens to be cheapest or most familiar.
Key takeaways
- 16S profiles who is there cheaply; shotgun adds what they can do, at higher cost.
- Amplicon resolution usually stops at genus; shotgun can reach species and strain.
- 16S function is inferred from taxonomy; shotgun measures gene content directly.
- The right choice depends on your question, budget, and sample type — not on which is “better”.
Two ways to read a microbiome
A microbiome sample — a gram of soil, a millilitre of gut content, a swab, a litre of seawater — contains DNA from hundreds or thousands of organisms. Both methods sequence that DNA, but they aim at different targets. Amplicon sequencing amplifies and reads one marker gene shared by the taxa of interest. Shotgun metagenomics fragments and sequences all the DNA in the sample. That single difference cascades into every practical trade-off below, and it is the reason there is no universally correct answer.
What each method sequences
In 16S rRNA sequencing, PCR primers target hypervariable regions (commonly V3–V4) of the bacterial and archaeal 16S ribosomal RNA gene. The gene is present in essentially all prokaryotes and contains regions conserved enough to prime against and variable enough to tell taxa apart. For fungi, the equivalent marker is the ITS region; for some eukaryotes, 18S. Because only a short marker is amplified, sequencing is cheap and shallow depth goes a long way.
Shotgun metagenomics skips amplification of a single gene and sequences random fragments of every genome present — bacterial, archaeal, viral, fungal, and host. Instead of one marker, you recover reads from across whole genomes, which is what makes functional analysis and strain-level resolution possible. It also means most of your reads may come from the most abundant organisms, so rare members need much deeper sequencing to detect.
Taxonomic resolution
Amplicon data are typically resolved to genus, and sometimes species, depending on the marker region and the reference database. Modern amplicon workflows use exact sequence variants (ASVs) rather than clustered OTUs, which improves reproducibility across studies, but the underlying limit is the short marker: two species with near-identical 16S regions are hard to separate. Taxonomy is assigned against curated references such as SILVA or the Genome Taxonomy Database (GTDB).
Shotgun data can reach species and even strain resolution because reads come from across the genome, not one conserved gene. Tools such as Kraken2 classify reads against large genome databases, and marker-based profilers like MetaPhlAn estimate species-level relative abundance. If distinguishing closely related organisms matters — tracking a pathogenic strain, say — shotgun is usually the only option.
Function: inferred vs. measured
This is the sharpest distinction. With 16S you know who is present, and you can infer likely function using tools such as PICRUSt2, which predict gene content from taxonomy and reference genomes. Inference is useful for hypothesis generation but is exactly that — a prediction of metabolic potential based on what related organisms typically carry, not a measurement of the genes in your sample.
Shotgun metagenomics measures gene content directly. You can profile metabolic pathways with HUMAnN, detect antibiotic-resistance or virulence genes, assemble contigs with MEGAHIT or metaSPAdes, and bin them into metagenome-assembled genomes (MAGs) that reconstruct near-complete genomes of uncultured organisms. If your question is functional — what can this community do, not just who is in it — shotgun is the honest choice.
Cost, depth, and host DNA
Amplicon sequencing is markedly cheaper per sample and needs far less depth, which is why large surveys with many samples often use 16S: you can profile hundreds of samples affordably and still see community-level patterns. Shotgun costs more per sample and needs greater depth, especially to characterise rare members or to assemble genomes.
Sample type also intrudes. Host-associated samples — a tissue biopsy, a tumour, a blood sample — can be overwhelmingly host DNA, so a shotgun run may spend most of its reads sequencing the human genome rather than microbes, and host reads must be removed before analysis. Amplicon sidesteps this because the primers target microbial marker genes, though primer choice introduces its own biases. Neither method is bias-free; the goal is to choose the biases you can live with for your question.
The bioinformatics differ, too
The two methods diverge computationally as much as they do at the bench. A typical amplicon workflow denoises reads into ASVs with DADA2 inside a framework such as QIIME 2, assigns taxonomy against SILVA or GTDB, and then computes diversity metrics and differential abundance. It is comparatively light on compute.
A shotgun workflow does quality control and host removal, then either profiles reads directly (Kraken2, MetaPhlAn, HUMAnN) or assembles and bins them into MAGs with quality assessment by CheckM. It is heavier on compute, storage, and interpretation. In both cases, the statistics that follow — diversity, ordination, and differential abundance on sparse, compositional data — are where community-level questions are actually answered, and they reward careful biostatistics. We cover both routes across our environmental genomics and infectious-disease work.
Which should you choose?
A short decision guide. Choose 16S/amplicon when you have many samples, a limited budget, and a community-structure question — how does composition differ between groups, sites, or time points. Choose shotgun when you need species or strain resolution, direct functional or pathway information, recovery of genomes (MAGs), or detection of viruses and plasmids that a 16S primer will never see. Many well-designed studies use amplicon to survey broadly and cheaply, then apply shotgun to a focused subset for depth — the best of both, if the budget allows.
A note on eDNA and metabarcoding
Environmental DNA (eDNA) studies extend amplicon thinking to biodiversity: marker genes amplified from water, soil, or air detect which taxa are present without capturing or observing them. It is a powerful survey tool, but the honest framing is detection, not census — a positive signal means DNA was present, and relative read counts are not absolute abundances. The same statistical care that applies to 16S applies here.
Study design and controls
No amount of clever analysis rescues a poorly designed microbiome study, and community data are unusually easy to bias. Two controls belong in every run. Negative (blank) controls — extraction and PCR blanks carried through the whole protocol — reveal reagent and lab contamination, which matters enormously in low-biomass samples like water, air, or tissue, where background can rival the real signal. Mock communities, samples of known composition, verify that your pipeline recovers what it should and expose primer or database biases before they mislead you.
Design choices then shape what the statistics can say. Randomise samples across extraction batches and sequencing lanes so a batch effect cannot masquerade as biology, and record those batches so the model can account for them. Plan enough biological replication for the comparison you care about, because community differences are noisy. Finally, respect the nature of the data at analysis time: microbiome counts are compositional — they sum to a fixed sequencing depth — so naive comparisons mislead, and either rarefaction or compositional-aware methods are needed. Getting design and controls right is precisely where careful biostatistics turns raw reads into conclusions that survive review.
Conclusion
The 16S-versus-shotgun decision is not about which method is superior; it is about matching the method to the question, the samples, and the budget. Amplicon answers “who is there” cheaply and at scale; shotgun answers “who is there and what can they do” at higher cost and complexity. Decide what you actually need to conclude before you sequence, and the choice usually makes itself. If you’d like a recommendation for your specific study, our metagenomics and microbiome service scopes the method and runs the full analysis.