DADA2: High-resolution sample inference from Illumina amplicon data¶
Why this mattered¶
Before DADA2, marker-gene microbiome studies commonly compressed sequencing reads into operational taxonomic units, often clustered at 97% similarity. That practice made noisy Illumina amplicon data tractable, but it also blurred biologically meaningful sequence differences and made results less portable across studies. Callahan et al. reframed the problem: instead of clustering reads into approximate taxa, model the sequencing errors and infer the exact amplicon sequences present in each sample. The paper showed that DADA2 could resolve variants differing by a single nucleotide while reducing spurious calls in mock communities.
The shift mattered because it turned amplicon sequencing from a coarse community-fingerprinting method into a higher-resolution assay for reproducible sequence variants. Exact amplicon sequence variants could be compared across datasets without redefining study-specific OTU clusters, enabling finer ecological, clinical, and evolutionary questions: strain-like variation, subtle community shifts, and longitudinal dynamics that had often been hidden by clustering. Its demonstration in vaginal microbiome samples, where previously undetected Lactobacillus crispatus variants were resolved, illustrated that the gain was not only technical but biologically interpretable.
DADA2 also helped set the methodological foundation for modern microbiome pipelines. Alongside related denoising approaches, it moved the field toward ASV-based analysis in tools such as QIIME 2 and Bioconductor workflows, improving reproducibility and making large cross-study comparisons more meaningful. Subsequent microbiome association studies, meta-analyses, and clinical ecology work inherited this higher-resolution unit of observation, making DADA2 a paradigm-shifting paper not because it introduced a new sequencing platform, but because it changed what Illumina amplicon reads were understood to contain.
Abstract¶
(no abstract available)