PDFBench (De Novo Protein Design from Function)
Unified benchmark for de novo protein design from functional descriptions. Supports description-guided and keyword-guided design tasks. 16-22 metrics across plausibility, foldability, language alignment, similarity, novelty, and diversity.
Composite
68.9
Experimental validation
None
Stages
Hit ID
Modalities
protein_structuretextprotein_sequence
Task types
protein_designtext_guided_generation
Size
description: Multiple evaluation tasks
-22_metrics: 16
-22_metrics: 16
License
Unknown
First release
2025-05
Last updated
2025-05
Official site
Leaderboard
→ leaderboard
Dataset
→ dataset
Code / GitHub
→ repository
HuggingFace
→ HF
Paper
PDFBench: A Benchmark for De novo Protein Design from Function · · 2025 · paper · 3 citations
Flags
text_guided_designmulti_metric
Experts
—
Groups
—
Hosted by
—
Related benchmarks
Rubric (7-criterion)
rigor
4
coverage
4
maintenance
2
adoption
2
quality
4
accessibility
4
industry_relevance
4
Notes
First unified benchmark for function-guided protein generation. Addresses comparison challenges from proprietary datasets and inconsistent metrics. Inter-metric correlation analysis provides guidelines for metric selection.