Beta diversity helps you compare differences between samples. So if alpha diversity is about what’s inside a single sample, beta is about how different two samples are from each other. Whether you’re comparing treatment groups, timepoints, or environments, beta diversity gives you insight into how microbial communities shift. Cosmos-Hub calculates beta diversity using relative abundance at the species level, which makes comparisons consistent across all your samples.

❓“I want to compare both what species are there and how abundant they are.”

➡ Use: Bray-Curtis Distance (Recommended)
  • Accounts for both the species present and their relative abundances.
  • Great if you’re studying treatments or interventions that affect microbial load.
  • Helps detect even subtle shifts in species proportions (not just who’s there).
  • Ideal for biologically meaningful comparisons where abundance matters.
🧠 Takeaway: Bray-Curtis gives you a score from 0 (identical samples) to 1 (totally different). For example, if two samples have the same species but wildly different proportions (like 90% vs. 10% E. coli), Bray-Curtis picks up that difference. Jaccard wouldn’t. 🔬 Example: If healthy gut samples cluster tightly (i.e., look similar) while diseased ones scatter across the PCoA plot, it suggests that disease increases variation in community composition.

❓“I only care about whether species are present, not how many.”

➡ Use: Jaccard Distance
  • Looks only at presence or absence of species. Ignores how abundant they are.
  • Great for studying colonization patterns or species co-occurrence.
  • Useful when abundance data is noisy or not biologically meaningful.
  • Helps answer questions like: “Do these two environments host the same species?”
🧠 Takeaway: Two samples can have the same Jaccard distance whether a species is at 1% or 99%. If the species are there, they’re counted. No more, no less.

📊 How Do I Interpret the Results?

Cosmos-Hub visualizes both Bray-Curtis and Jaccard using PCoA (Principal Coordinates Analysis) plots.
  • Distance between dots = how different the samples are
  • Tight clusters = similar communities
  • Wide spread = more variation
  • Group separation = distinct community types (e.g., treated vs. untreated)

🧭 Which One Should I Use?

Best practice is to report p-values for all metrics (e.g., Bray-Curtis p=0.001, Jaccard p=0.042). But if only plotting one metric in your paper, follow these guidlines:
GoalBest Metric
Compare species + abundance (quantitative)Bray-Curtis
Compare species presence only (qualitative)Jaccard
Get the full pictureUse both

✔️ Bray-Curtis:

  • You’re analyzing treatment effects or dose-response changes
  • Abundance differences are biologically meaningful
  • Studying metabolic activity or disease progression

✔️ Jaccard:

  • You’re focusing on who’s there, not how much
  • Abundance is unreliable or irrelevant
  • Looking at biogeography or species colonization

✔️ Both:

  • You want to distinguish species turnover from abundance shifts
  • You’re publishing and want reviewers to see both angles
  • You need complementary insights (structure and membership)

📈 What About Statistical Testing?

Both Bray-Curtis and Jaccard can be analyzed using PERMANOVA (Permutational Multivariate Analysis of Variance):
  • Tests if microbial communities differ significantly between groups
  • Outputs include F-statistics and p-values
  • Helpful for validating patterns seen in your visualizations

✅ Cosmos-Hub Recommends…

Start with Bray-Curtis. It’s the most biologically informative metric for most use cases in microbiome research. But don’t hesitate to layer on Jaccard if your study asks broader questions about community membership.