Protein-Ligand Benchmark Demo

Status: Demo-only benchmark walkthrough Device: CPU-compatible

This walkthrough demonstrates the retained protein-ligand benchmark owners in explicit demo mode. The shipped script uses mock CrossDocked-style complexes and retained heuristic drug-likeness scoring; it should not be read as a benchmark-grade public CrossDocked runtime.

Files

• Python script: protein_ligand_benchmark_demo.py
• Jupyter notebook: protein_ligand_benchmark_demo.ipynb

Run It

python examples/generative_models/protein/protein_ligand_benchmark_demo.py
jupyter lab examples/generative_models/protein/protein_ligand_benchmark_demo.ipynb

What This Demo Actually Uses

• CrossDockedDataset with metadata={"demo_mode": True, ...}
• ProteinLigandBenchmarkSuite(..., demo_mode=True)
• ProteinLigandCoDesignBenchmark(..., demo_mode=True)
• DrugLikenessMetric in explicit demo mode plus retained validity and affinity helpers
• typed molecular constraints configured through ChemicalConstraintConfig under an explicit extensions={...} bundle

from artifex.generative_models.core.configuration import ChemicalConstraintConfig

extensions={
    "chemical": ChemicalConstraintConfig(name="chemical_constraints")
}

Why It Is Demo-Only

• the dataset owner still generates mock protein-ligand complexes instead of loading benchmark-grade CrossDocked assets
• drug-likeness scoring is retained heuristic logic, not a chemistry-backend benchmark path
• the page teaches explicit demo opt-in rather than a supported public benchmark command surface

Use This When

Use this pair when you want to inspect the retained protein-ligand benchmark interfaces, prototype against the demo batch shapes, or replace the demo data and heuristics with your own benchmark-grade backend from Python.