DynBench3D, a Web-Resource to Dynamically Generate Benchmark Sets of Large Heteromeric Protein Complexes

Area of Science:

• Structural Biology
• Computational Biology
• Biochemistry

Background:

• Multi-protein machines perform essential cellular functions, but their atomic-level structures are often unknown.
• Computational methods integrate diverse structural data to model large molecular machines.
• A lack of gold-standard benchmark datasets hinders the evaluation of these modeling methodologies for complexes.

Purpose of the Study:

• To develop a strategy for dynamically generating non-redundant sets of 3D heteromeric complexes with three or more components.
• To create a reliable benchmark for assessing computational methods in structural biology.

Main Methods:

• Developed a dynamic generation strategy for non-redundant 3D heteromeric complexes.
• Varied sequence identity and component overlap thresholds to define complex redundancy.
• Identified 495 unique target complexes using a 20% identity threshold and <0.5 component overlap.

Main Results:

• Generated a non-redundant set of 495 unique 3D heteromeric complexes.
• Provided template subcomplexes for each target complex to facilitate modeling exercises.
• The interactive resource is available at https://DynBench3D.irbbarcelona.org.

Conclusions:

• The developed strategy provides a crucial resource for benchmarking and advancing computational methods for modeling macromolecular assemblies.
• This work will accelerate the development and progress assessment of novel methodologies in structural biology.
• Facilitates the understanding of complex protein machinery at an atomic level.