Exploring the effects of sparse restraints on protein structure prediction

Area of Science:

• Computational Biology
• Biophysics
• Structural Biology

Background:

• Accurate protein structure prediction is challenging due to the vast conformational space.
• Distance restraints and inter-residue contacts can reduce this search space.
• Previous work suggested ~1 contact per 12 residues might suffice for fold-level accuracy.

Purpose of the Study:

• To investigate the minimum number and type of contacts required for accurate protein structure prediction.
• To evaluate the efficacy of sparse contact maps using coarse-grained models and molecular dynamics.
• To understand the relationship between contact map sparsity and protein folding success.

Main Methods:

• Utilized coarse-grained, structure-based models.
• Performed molecular dynamics simulations.
• Generated sparse contact maps for 15 proteins of diverse lengths and topologies.

Main Results:

• A small fraction (5-10%) of native contacts, with perfect secondary structure information, is sufficient for correct protein folding.
• Not all sparse maps are equivalent; specific types are more predictive.
• Long-range contacts are more informative, especially for alpha and alpha-beta proteins, less so for beta-proteins.

Conclusions:

• Structure-based models effectively assess the utility of structure prediction restraints.
• Predictive sparse maps can be generated by consensus or by selecting well-distributed contacts.
• Future models can incorporate force-field interactions, secondary structure errors, and noise.