Diversity comparison

Compare within-sample and between-sample diversity across groups, with the sequencing-depth confound kept visible.

Research question

Do communities differ in alpha diversity or overall composition between groups, and is any separation statistically supported rather than a depth artifact?

Who this is for

Microbiome researchers comparing treatment, site, or condition groups
Ecologists summarizing community structure and stability
Teams needing assumption-aware diversity statistics

Data requirements

Data	Required	Purpose
Feature table (taxa × samples)	Yes	Alpha and beta diversity computation
Sample metadata with a group column	Recommended	PERMANOVA group separation and stratified views

Workflow

Upload feature table + metadata → Alpha diversity with depth context + rarefaction
  → Alpha diversity group significance (optional)
  → Beta diversity (Bray-Curtis or Jaccard) + PCoA ordination
  → PERMANOVA + PERMDISP group separation (optional)
  → AI interpretation

Step 1 — Alpha diversity and rarefaction

Run the Diversity analysis to compute Shannon, Simpson, observed features, and Chao1 per sample, reported alongside sequencing depth so separation is never presented without its confound. A rarefaction curve (multinomial subsampling across log-spaced depths) shows whether samples were sequenced deeply enough for the alpha metrics to be trusted.

Step 2 — Alpha diversity group significance

When a grouping column is supplied, a Kruskal-Wallis test (with BH-corrected pairwise Wilcoxon tests between groups) reports whether alpha diversity itself differs between groups — a companion to PERMANOVA, which only tests beta-diversity separation.

Step 3 — Beta diversity, PCoA, PERMANOVA, and PERMDISP

Choose Bray-Curtis or Jaccard for beta-diversity distances and review the PCoA ordination. When a grouping column is supplied, PERMANOVA tests group separation and PERMDISP (betadisper) tests whether that separation could instead reflect unequal within-group dispersion — both are associational results, not causal ones.

Expected outputs

Per-sample alpha-diversity metrics and cohort means with depth context
Rarefaction curves for sequencing-depth adequacy
Kruskal-Wallis and pairwise Wilcoxon results for alpha-diversity group differences
Beta-diversity distances (Bray-Curtis or Jaccard) and PCoA coordinates with variance explained
PERMANOVA test statistic and p-value, plus PERMDISP F-statistic and p-value, for group separation