Almost every microbiome project begins with the same decision: sequence one marker gene, or sequence everything. Get it wrong and you either spend far more than you needed to, or you finish the project unable to answer the question you started with.
This article sets out what each method actually measures, what it costs, and how to choose between them.
The two approaches
16S rRNA amplicon sequencing targets a single gene present in all bacteria and archaea. The 16S gene contains nine hypervariable regions (V1–V9) flanked by conserved stretches. You PCR-amplify one or two of those regions with universal primers and sequence the product. Every read comes from the same gene, so you are counting taxa.
Shotgun metagenomics fragments and sequences all DNA in the sample. Reads come from every gene of every organism present, including bacteria, archaea, fungi, viruses and any host DNA. You are sampling the entire genetic content of the community.
What each one can tell you
Taxonomic resolution
16S reliably resolves to genus. Species-level assignment is possible sometimes but frequently unreliable, because the amplified region may be identical between closely related species. If distinguishing E. coli from Shigella matters to your conclusion, 16S will not do it.
Shotgun routinely reaches species and, with adequate depth, strain level. For clinical questions — where a pathogenic strain and a commensal of the same species behave completely differently — this distinction is often the entire point.
Who is present, versus what they can do
This is the fundamental difference. 16S tells you who is there. Shotgun tells you what the community is capable of.
Only shotgun data gives you direct evidence of antimicrobial resistance genes, virulence factors, metabolic pathways, biosynthetic gene clusters or carbohydrate-active enzymes. Tools such as PICRUSt2 attempt to infer function from 16S by mapping taxa to reference genomes, but this is inference from a database, not measurement. It performs acceptably for well-characterised environments like the human gut and poorly for soil, marine or other under-referenced communities. It cannot detect a resistance gene on a plasmid, because plasmid content varies between strains sharing an identical 16S sequence.
Non-bacterial members
16S primers amplify bacteria and archaea. Fungi need ITS sequencing; viruses have no universal marker gene at all. If your question involves the virome or mycobiome, shotgun is the only single-assay option.
The biases you should know about
Neither method is neutral, and the biases differ.
16S carries PCR bias. Universal primers are not universal — they match some taxa better than others, so relative abundances are distorted in ways that depend on which primer set you chose. Different hypervariable regions give measurably different community profiles from the same sample, which is a major reason 16S studies using different regions are hard to compare. On top of that, 16S copy number varies between organisms, from one to more than fifteen copies per genome, so read counts do not map cleanly onto cell counts without correction.
Shotgun carries host contamination and database bias. In a tissue or swab sample, host DNA can dominate, wasting most of your sequencing budget; low-biomass samples may yield very little microbial signal. And classification depends on reference databases, so novel or under-represented organisms are misassigned or missed — a serious limitation in environmental work.
Cost and depth
16S is cheap. You need perhaps 10,000–50,000 reads per sample, so hundreds of samples fit comfortably on one run. That makes it the practical choice for large cohorts, longitudinal designs and screening.
Shotgun typically needs 5–20 million reads per sample for solid taxonomic profiling, and considerably more for assembly and recovery of metagenome-assembled genomes. Costs are perhaps an order of magnitude higher per sample, and the analysis is much heavier — assembly and binning need substantial memory and compute, where a 16S analysis runs comfortably on a laptop.
How to choose
Choose 16S when you have many samples and a limited budget; the question is comparative (does community composition differ between groups?); you are screening before committing to deeper work; or you are working with low-biomass samples where host DNA would swamp a shotgun run.
Choose shotgun when you need species or strain resolution; the question is functional — resistance genes, pathways, virulence; viruses or fungi matter; or you intend to recover genomes from the community.
A common and sensible design uses both: 16S across the full cohort to find where the interesting differences are, then shotgun on a selected subset to establish what is happening functionally. This is usually a better use of a fixed budget than shallow shotgun on everything.
Analysis in outline
A 16S pipeline today typically uses DADA2 or Deblur to infer exact amplicon sequence variants rather than clustering at 97% into OTUs — ASVs are reproducible between studies in a way OTUs are not. Taxonomy is assigned against SILVA, Greengenes2 or RDP, then alpha and beta diversity and differential abundance follow.
A shotgun pipeline begins with quality trimming and host read removal, then either read-based profiling (Kraken2 with Bracken, or MetaPhlAn) or assembly with metaSPAdes or MEGAHIT followed by binning into MAGs. Functional annotation runs against databases such as KEGG, CARD for resistance and CAZy for carbohydrate-active enzymes.
One thing both methods share
Sequencing gives relative abundance, not absolute. If one taxon blooms, every other taxon appears to decrease as a proportion even when its absolute cell count is unchanged. This compositional property invalidates many standard statistical tests, and ignoring it is one of the most common analytical errors in microbiome work. Use methods designed for compositional data — ALDEx2, ANCOM-BC or centred log-ratio transformation — or pair sequencing with qPCR or flow cytometry if you genuinely need absolute quantification.
