Lemon

Lemon mines three-dimensional features from macromolecular structures to generate standardized benchmarking datasets and support virtual screening and structural biology analyses.

Key Features:

• 3D feature mining: Extracts and organizes three-dimensional features from protein and ligand structures in the Protein Data Bank (PDB).
• DUBS integration: Integrates with the Directory of Useful Benchmarking Sets (DUBS) to produce standardized benchmarking sets.
• Text-based benchmark definition: Uses a text-based input format to define and generate custom benchmarking datasets.
• Benchmark generation speed: Generates a benchmark dataset via DUBS in under two minutes.
• Python-script extensibility: Provides a Python script interface for modifying and extending benchmark definitions and processing.
• Standardized dataset representation: Produces standardized representations intended to improve reproducibility and consistency in virtual screening benchmarking.

Scientific Applications:

• Virtual screening benchmarking: Supports creation and evaluation of benchmarking datasets for virtual screening methods in drug discovery.
• Structural biology analyses: Enables mining and analysis of 3D macromolecular features for structural biology studies.
• Method comparison and reproducibility: Facilitates standardized benchmarking to compare computational methods and enhance reproducibility.

Methodology:

Performs PDB data mining to access protein and ligand structures, uses a text-based input format and a Python script to define, generate, and standardize benchmarking datasets, and integrates with the DUBS framework (benchmark generation reported in under two minutes).