Princeton_TIGRESS: protein geometry refinement using simulations and support vector machines.

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• Protein structure refinement aims to improve predicted models towards native structures.
• Many existing computational methods often degrade model quality rather than improve it.
• Previous CASP (Critical Assessment of protein Structure Prediction) experiments highlight challenges in refinement.

Purpose of the Study:

• To present and evaluate the Princeton_TIGRESS protein structure refinement protocol.
• To assess the performance of Princeton_TIGRESS on historical CASP refinement targets.
• To provide a robust and accessible tool for improving protein model accuracy.

Main Methods:

• Development and application of the Princeton_TIGRESS refinement protocol.
• Benchmarking on CASP 7, 8, 9, and 10 refinement targets.
• Testing on models generated by top-performing structure prediction servers.
• Evaluation of protocol robustness using different random seeds.

Main Results:

• Princeton_TIGRESS increased GDT_TS (Global Distance Test Total Score) in 76% of cases.
• An average improvement of 0.83 GDT_TS points per structure was observed.
• The method demonstrated robustness across different random seeds and input models.

Conclusions:

• Princeton_TIGRESS consistently improves protein model accuracy.
• The protocol offers a reliable solution for bridging the gap between predicted and native structures.
• The refinement protocol is available as a free web server for the scientific community.