DynaBench: Dynamic data for the docking benchmark.

Area of Science:

• Biochemistry
• Structural Biology
• Computational Biology

Background:

• Protein-protein interactions are crucial for cellular functions like transport and signaling.
• Current structural modeling tools (e.g., AlphaFold) provide static representations, neglecting critical interface flexibility.
• Understanding protein interface dynamics is essential for accurate functional insights.

Purpose of the Study:

• To introduce DynaBench, a comprehensive benchmark for protein-protein interface dynamics.
• To provide a large-scale dataset of molecular dynamics (MD) simulations for protein complexes.
• To facilitate advancements in computational modeling and analysis of protein assemblies.

Main Methods:

• Performed extensive MD simulations on over 200 protein-protein complexes from the Docking Benchmark 5.5.
• Generated three 100 ns long trajectory replicas for each complex.
• Made all simulation data publicly available through the MDposit platform within the Molecular Dynamics Data Bank (MDDB).

Main Results:

• Generated a substantial dataset of protein complex dynamics, capturing interfacial flexibility.
• Established a valuable resource for training machine learning models for protein structure prediction.
• Enabled the exploration of novel accuracy metrics for evaluating protein complex models.

Conclusions:

• DynaBench addresses the limitation of static models by providing crucial interface dynamics data.
• The benchmark serves as a key resource for computational structural biology and drug design.
• Publicly accessible data promotes further research in protein complex modeling and functional analysis.