Dockground scoring benchmarks for protein docking.

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Protein docking requires global scans followed by re-ranking with accurate scoring functions.
• Scoring function development often uses benchmark sets (decoys) representing global scan outputs.
• Existing benchmarks may not fully reflect real-world docking challenges.

Purpose of the Study:

• To create novel protein-protein scoring benchmark sets for the Dockground resource.
• To provide a resource for developing and testing protein-protein scoring approaches.
• To generate realistic docking decoys that mimic real-world applications.

Main Methods:

• Generated two protein-protein scoring benchmark sets based on Dockground benchmarks (unbound docking benchmark 4 and model-model benchmark 2).
• Designed decoys to represent realistic docking scenarios, defining correct predictions as near-native structures.
• Minimized biases by reducing prediction clustering and structural quality disparities between near-native and non-native matches.
• Characterized benchmark sets by source organism and protein complex function.

Main Results:

• Developed two new, user-friendly protein-protein scoring benchmark sets.
• The sets offer a concise and nonredundant representation of global docking scan outputs.
• Decoys are designed to be applicable to scoring function development, avoiding non-scoring related approaches.

Conclusions:

• The new Dockground benchmark sets provide a valuable resource for the research community.
• These datasets facilitate the development and validation of advanced protein-protein scoring methodologies.
• The resource supports the advancement of computational drug discovery and protein interaction studies.