Structures · Database
CATH
A structural hierarchy classifying protein domains and their evolutionary relationships.
What CATH is
CATH is a classification database for protein domains, created in the mid-1990s by Christine Orengo, Janet Thornton, David Jones, and colleagues. Maintained by the Orengo group at University College London, it provides structural and evolutionary organisation for experimentally determined protein structures derived from the Protein Data Bank (PDB).
Because protein three-dimensional structures are generally more conserved than primary sequences, structural comparisons can identify distant evolutionary relationships that sequence alignments miss. CATH addresses this by partitioning protein chains into structural domains—independently folding units—and organising them into an evolutionary hierarchy.
The resource is linked with Gene3D and related initiatives like The Encyclopedia of Domains (TED) to extend structural annotations to millions of uncharacterised sequences and predicted models, aiding researchers in protein function prediction and evolutionary analyses.
How it is organised
Experimentally determined structures from the PDB are separated into individual polypeptide chains, which are then segmented into domain units using a combination of automated procedures and manual curation. These domains are assigned into a four-tier hierarchy: Class (C), based on secondary structure composition (such as all-alpha, all-beta, mixed alpha-beta, or low secondary structure); Architecture (A), reflecting secondary structure arrangement in space; Topology or fold (T), describing spatial arrangement and connectivity; and Homologous Superfamily (H), grouping domains that share demonstrable common evolutionary ancestry.
Superfamilies are further partitioned into structural clusters and Functional Families (FunFams), which group domains sharing specific functions. Gene3D builds hidden Markov models (HMMs) for these clusters to map homologous domains onto sequence databases. Function and conserved residues within FunFams are evaluated using alignment conservation metrics like Scorecons, mapping conserved positions directly onto representative structures.
Key facts
- Classification hierarchy
- Class (C), Architecture (A), Topology (T), Homologous superfamily (H)
- Domains in CATH-Plus 4.4.0
- 601,493
- Superfamilies in CATH-Plus 4.4.0
- 6,573
- Annotated PDBs in CATH-Plus 4.4.0
- 150,885
- Gene3D protein sequences
- 82,665,384
- Gene3D domain predictions
- 151,013,797
- Inputs
- Protein amino acid sequences or PDB coordinates
What you can use it for
- Sequence-based structural domain identification
- Querying a protein sequence against CATH to identify matching domain boundaries, 3D structures, and homologous superfamilies.
- Protein function prediction
- Assigning putative Enzyme Commission (EC) numbers and Gene Ontology (GO) terms to uncharacterised sequences via matches to specific Functional Families (FunFams).
- Functional residue and active site analysis
- Identifying conserved functional residues using FunFam multiple sequence alignments and Scorecons conservation scores mapped to representative domain coordinates.
- Investigating domain evolution and structural diversity
- Exploring sequence clusters, structural similarity groups, and shared core geometries within homologous superfamilies.
Strengths and limits
Good at
- Hierarchical grouping connects spatial architecture with evolutionary lineage down to functional family sub-clusters.
- Integrated Gene3D sequence models allow structural domain annotations to be assigned directly to raw protein sequences.
- FunFam function prediction methodology achieves high accuracy rankings in Critical Assessment of protein Function Annotation (CAFA) evaluations.
- Provides daily snapshot updates (and intermediate CATH-B annotations) alongside formal annual releases to minimise latency relative to the PDB.
Watch out for
- Official releases occur roughly once every 12 months, creating a lag for fully validated annotations of recent PDB structures.
- A small number of superfamilies display extreme sequence, structural, and functional divergence, complicating simple structural comparisons.
Questions
Can I search CATH without writing code?
CATH is catalogued here but not yet queried live, so the agent can point you to it but cannot search it for you yet.
Can I reuse data from CATH?
CATH publishes its own terms of use; read them on https://www.cathdb.info before reusing its records.
What do the letters in CATH stand for?
They represent the four primary levels of its classification hierarchy: Class (C), Architecture (A), Topology/fold (T), and Homologous superfamily (H).
How does CATH handle newly deposited structures between major releases?
It provides daily snapshot updates and CATH-B, which supplies preliminary domain boundaries and superfamily classifications for newly deposited PDB structures.
What is the role of FunFams in CATH?
Functional Families (FunFams) are sub-classifications of homologous superfamilies that group domains sharing the same functional annotations, providing Enzyme Commission (EC) numbers and Gene Ontology (GO) terms for queries.
Can CATH be queried using only a protein sequence?
Yes, user protein sequences can be submitted to search against CATH-Gene3D models to identify matching superfamilies, FunFams, and functional annotations.
How to cite it
Waman VP, Bordin N, Lau A, Kandathil S, Wells J, Miller D, Velankar S, Jones DT, Sillitoe I, Orengo C. CATH v4.4: major expansion of CATH by experimental and predicted structural data. Nucleic Acids Res. 2025 Jan. doi: 10.1093/nar/gkae1087
Written from CATH's own website and reference material. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.