DisProt: Manually Curated Intrinsically Disordered Protein Annotations

3,199 protein entries with manually curated annotations for intrinsically disordered regions and experimental evidence. The dataset was created by LiteFold and last updated on HuggingFace in May 2026. It includes a deterministic split of 2,875 rows for training and 324 for testing.

Use Cases

• Train disorder prediction models based on manually curated protein disorder annotations.
• Benchmark computational methods against experimental evidence for protein disorder.
• Analyze functional annotations associated with intrinsically disordered regions.

Strengths

• 3,199 protein entries provide a substantial corpus for analysis.
• Data is manually curated, which likely indicates high annotation quality.
• Includes a deterministic train/test split of 2,875 and 324 rows respectively.

Limitations

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

Provenance

Source

DisProt database

Collection Method

Manually curated annotations

Freshness

Last updated 2026-05-27 12:56:17; freshness should be verified.