ProteinFamilyTree
Also: ZooFold, call whose version of a protein is closest to ours, then see the letters and the shapes.
How much does one protein change across the tree of life? Pick a protein family below to see its versions in a ladder of model organisms (bacteria → yeast → fly → fish → mouse → human), each compared with the human one twice: letter by letter (how many aligned letters match) and by shape (how far apart the AlphaFold predictions sit, on the residues both are sure of). Nothing is installed and nothing about you is collected.
This is the tool. A native version, when it ships, adds an AI-narrated read-out, bookmarks, and a classroom mode; it never gates the web.
Pick a protein family
Family ladder
Each model organism that carries this protein, in teaching order (E. coli at the top, human at the bottom). This is a teaching order, not a phylogenetic distance. Select any organism to load its structure. Beside each: its UniProt accession and the share of aligned letters that match ours; on the right, how far its predicted shape sits from ours (Cα RMSD, shorter = closer).
Shape distance from human
The Cα RMSD after the best rigid fit of each organism's predicted shape onto the human one, over the residues that line up in the sequence alignment and that both AlphaFold models are sure of (pLDDT 70 or more). A bar marked two pieces matches closely part by part, but its parts sit differently relative to each other; not compared means too few sure residues to fit.
Compare structures
Methods & limits
- The ladder is a teaching order, not a phylogeny. Organisms are laid out E. coli → human as a familiar teaching sequence. It is not a phylogenetic tree and the spacing is not an evolutionary distance — this tool does no tree inference, no molecular clock, and no bootstrap.
- Every protein is looked up, never typed. Each organism's version is the one reviewed (Swiss-Prot) UniProt entry for its gene in that organism; its organism and protein name are stored with it and checked when the site is built. Where an organism has no reviewed entry for a family, it is left out rather than guessed.
- Letters: a global sequence alignment. "% letters" is the share of aligned positions where the two sequences carry the same amino acid (Needleman–Wunsch, BLOSUM62, gap open 10 / extend 0.5; overhanging ends are not counted).
- Shape: a rigid fit on aligned, sure residues. The Cα RMSD is a Kabsch superposition over the residues the alignment pairs up and that both AlphaFold models are sure of (pLDDT 70+). It is not a structural aligner (no TM-align). When one rigid fit is poor but two parts each fit well, it says so instead of giving one misleading number.
- Structures are AlphaFold predictions. Each shape is a predicted model from the AlphaFold DB, not an experimental structure. A prediction can sit a protein's movable parts in one arrangement of several.
- The numbers are worked out ahead of time, from stored data. The letters, the shapes and the verdict are computed from the stored sequences and models (snapshot date shown under the verdict) and re-checked on every build; any number in the text beside them comes from the same calculation.
About this tool
Structures come from the AlphaFold Protein Structure Database (EMBL-EBI / Google DeepMind). They are fetched directly by your browser and rendered with Mol* via PDBe Mol*, self-hosted here — no third-party CDN, no tracking, no server. Protein entries and sequences come from UniProt. Please cite the AlphaFold DB and UniProt when you use this tool. It is part of the free, non-profit Fold Commons project.