Atomic-accuracy prediction of protein loop structures through an RNA-inspired Ansatz

Area of Science:

• Computational Biology
• Structural Biology
• Biophysics

Background:

• Predicting biopolymer structure from sequence alone is challenging, especially for protein loops longer than 10 residues.
• Existing methods like kinematic inversion closure (KIC) Monte Carlo simplify protein models, limiting systematic exploration of all-atom configurations.
• Accurate loop prediction is crucial for comparative modeling and protein design.

Purpose of the Study:

• Introduce a novel all-atom enumeration strategy for protein structure prediction.
• Evaluate the performance of the stepwise assembly (SWA) protocol against established methods.
• Demonstrate SWA's capability in modeling challenging and unusual protein loop structures.

Main Methods:

• Developed the stepwise assembly (SWA) protocol based on residue-by-residue enumeration and dynamic programming within the Rosetta framework.
• Applied SWA to predict conformations of protein loops, including those with irregular or unusual lengths.
• Compared SWA's accuracy against KIC Monte Carlo on established benchmarks and blind tests.

Main Results:

• SWA achieved sub-Angstrom accuracy for 19 of 20 loops in a benchmark, surpassing KIC's 14 of 20 accuracy.
• Successfully modeled unusual loop structures up to 24 residues, including cis-Pro turns and loops traversing side-chain tunnels.
• Achieved sub-Angstrom accuracy in five blind tests, including a protein/RNA binding interface model.

Conclusions:

• All-atom enumeration via SWA provides a systematic and accurate approach to ab initio protein structure modeling.
• SWA leverages high-performance computing and realistic energy functions for consistent atomic accuracy.
• Limitations are primarily due to the accuracy of the underlying Rosetta all-atom energy function.