Two modes depending on the dataset

The button shows Kingdom for mixed-kingdom datasets (Bacteria + Archaea + Eukaryota) and Phylum for single-kingdom datasets (Bacteria-only, Eukaryota-only, or Archaea-only). In both cases the underlying question is the same: where is this protein family most prevalent?

Kingdom mode (mixed datasets)

For each Pfam, we count how many organisms in each kingdom carry it (copy count > 0), then divide by the total number of organisms from that kingdom in scope:

• Bacteria — 9,091 organisms
• Archaea — 303 organisms
• Eukaryota — 2,765 organisms

This normalization ensures bacteria don't dominate by sheer numbers.

• Rare — max prevalence across all three kingdoms < 5%.
• Universal — 2nd-highest kingdom prevalence is ≥ 60% of the highest (e.g. ribosomal proteins).
• Bacteria / Archaea / Eukaryota — the kingdom with the highest prevalence when it is ≥ 1.67× the next.

Phylum mode (single-kingdom datasets)

For single-kingdom datasets, coloring by kingdom would be uninformative — every pfam is by definition most prevalent in that one kingdom. Instead, the same prevalence logic is applied one level lower, across phyla within that kingdom:

• Rare — max prevalence across all phyla in this kingdom < 5%.
• Widespread — 2nd-highest phylum prevalence is ≥ 60% of the highest (broadly distributed, e.g. core housekeeping genes in bacteria).
• Phylum name — the phylum with the highest prevalence when it dominates.

Example for Bacteria-only datasets: a pfam in 80% of Pseudomonadota but only 10% of other phyla → colored as Pseudomonadota.

Precomputation

All classifications are precomputed by precompute_kingdoms.php and stored in umap2.pfam_kingdom_profile. The color map is fetched once per session when this mode is first activated, then cached client-side.