Dimensionality reduction & clustering / PCoA
Your data stays in your browser: files are read and processed on your device, never uploaded to any server, and cleared when you close the page. Free, no sign-in.
PCoA places samples in a plane so that their distances are preserved as far as possible: closer points are more similar under the chosen distance. The percentage on each axis is its eigenvalue as a share of all positive eigenvalues.
Community data usually use Bray–Curtis (abundance-weighted) or Jaccard (presence / absence). With Euclidean distance, PCoA equals an unscaled PCA.
Non-Euclidean distances such as Bray–Curtis give negative eigenvalues: by default they are set to 0 and their size is reported in a warning; if they are large, try the Lingoes correction or NMDS.
Ellipses are 95% normal confidence ellipses of each group’s coordinates and only describe within-group spread; test group differences with PERMANOVA or ANOSIM.
You can also upload a distance matrix directly (e.g. UniFrac): first column = sample names, then one column per sample in the same order.
Gower J. C. (1966) Biometrika 53:325–338; Legendre & Legendre (2012) Numerical Ecology §9.3; Lingoes (1971) Psychometrika 36:195–203. Coordinates, eigenvalues and proportions match scikit-bio pcoa; distances match scipy.spatial.distance.pdist (Jaccard on presence / absence).
Up to 3,000 samples (above 500 samples only the first 10 axes are computed).
Send us your data and research question and you will receive a written plan within 1 working day: analysis steps, parameter rationale, deliverables and timeline. Quoted per project.