TEDLH: TED-to-CATH HMM Database for Protein Domain Analysis

A database of Hidden Markov Models (HMMs) for protein domains, mapped from the TED database to CATH structural classifications and filtered for a maximum pairwise sequence identity of 30%. Claudia Alvarez-Carreño authored this dataset, which was last updated on June 4, 2026. It contains 934,186 HMMs spanning 4,688 CATH superfamilies.

Use Cases

• Remote homology detection for protein sequences based on Hidden Markov Models.
• Classification of protein domains into structural superfamilies based on CATH mapping.
• Training or benchmarking protein language models based on a large set of filtered protein domain sequences.
• Comparative analysis of protein domain architecture across different structural classes (Mainly Alpha, Mainly Beta, Alpha Beta).

Strengths

• Contains 934,186 HMMs, providing a substantial resource for protein domain analysis.
• Organized into three distinct structural classes with specific counts: 212,345 HMMs for Class 1, 254,974 for Class 2, and 466,867 for Class 3.
• Maps to 50,167 CATH-PDB sequences, offering a direct link to experimentally determined protein structures.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific modeling tasks.

Provenance

Source

Derived from the TED database and mapped to CATH domains.

Collection Method

Filtered for a maximum pairwise sequence identity of 30%.

Freshness

Last updated 2026-06-04 20:15:54; freshness should be verified.