SCOPe: 15,176 Non-Redundant Protein Domains with Pairwise TM-Scores

University of Michigan researcher Chengxin Zhang compiled this dataset of 15,176 SCOPe protein domains from the ASTRAL subset with less than 40% sequence identity. It includes FASTA sequences, full atomic and CA-only PDB coordinate files, and pre-calculated pairwise TM-scores for structural alignment. The data originates from version 2.08 of the SCOPe database.

Use Cases

• Benchmarking protein structure alignment tools based on the provided pairwise TM-scores.
• Training machine learning models for protein structure prediction using the non-redundant set of domain structures.
• Analyzing structural relationships within protein fold families based on the SCOPe classification hierarchy included in the FASTA headers.

Strengths

• Contains 15,176 protein domains, providing a substantial collection for analysis.
• Offers a non-redundant set filtered to less than 40% sequence identity, reducing bias.
• Includes multiple data representations: sequences, full atomic coordinates, CA-only coordinates, and pre-computed similarity scores.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• Row count for the pairwise TM-score file is unknown, which may limit suitability assessment.

Provenance

Source

SCOPe database version 2.08, ASTRAL subset.

Collection Method

Domains were selected from the SCOPe database, sequences were converted from PDB files, and TM-scores were calculated using TMalign.