Differential abundance
Identify taxa that differ between two cohorts using compositional-data-aware methods, with prevalence and effect size reported alongside significance.
Research question
Which taxa are differentially abundant between two groups once compositionality and sparsity are accounted for, and how prevalent and how large is each difference?
Who this is for
- Microbiome researchers contrasting responder/non-responder or case/control cohorts
- Translational teams identifying candidate microbial biomarkers
- Analysts who need compositional methods rather than generic DE logic
Data requirements
| Data | Required | Purpose |
|---|---|---|
| Feature table (taxa × samples) | Yes | Per-feature abundance |
| Sample metadata with a group column | Yes | Defines the two cohorts to contrast |
Workflow
Upload feature table + metadata → Choose group column and two cohorts
→ Choose method: CLR + Mann-Whitney, or ANCOM-BC
→ Per-feature test + BH correction
→ AI interpretation
Step 1 — Configure the contrast
On the Differential analysis, choose the grouping column and (optionally) the two groups to compare. A prevalence filter removes features detected too rarely to test reliably. The shared rank selector collapses features to the same taxonomic rank used elsewhere in the study.
Step 2 — Compositional testing
Two methods are available: the default centered-log-ratio (CLR) transform with a rank-based Mann-Whitney test, or ANCOM-BC, which corrects for per-sample sampling fraction before a Welch t-test and reports a log-ratio effect size with a 95% confidence interval. Both are BH-corrected across features. The product does not apply generic transcriptomic differential-expression logic to microbial count tables.
Expected outputs
- Ranked features with effect size (CLR log-ratio, or ANCOM-BC log-ratio + 95% CI) and prevalence
- BH-corrected q-values and significance flags
- Per-group mean relative abundance for each feature