Improved PEP-FOLD Approach for Peptide and Miniprotein Structure Prediction

Area of Science:

• Computational biology
• Structural bioinformatics
• Biophysics

Background:

• Peptides and mini-proteins have significant biological roles.
• Accurate prediction of their 3D structures from amino acid sequences is crucial for research.
• Existing methods require improvement for large-scale applications.

Purpose of the Study:

• To introduce PEP-FOLD2, an enhanced coarse-grained approach for *de novo* peptide structure prediction.
• To compare PEP-FOLD2's performance against PEP-FOLD1 and the Rosetta program.
• To evaluate the accuracy and efficiency of predicting native or near-native conformations.

Main Methods:

• Utilized a benchmark set of 56 structurally diverse peptides (25-52 amino acids).
• Performed 600 simulations for each peptide using PEP-FOLD2, PEP-FOLD1, and Rosetta.
• Assessed model quality and the success rate of predicting near-native or native conformations.

Main Results:

• PEP-FOLD2 generated higher quality models compared to PEP-FOLD1.
• PEP-FOLD2 achieved a 95% success rate in generating near-native/native models, compared to 88% for Rosetta.
• Without experimental data, PEP-FOLD2 placed near-native/native models in the top five for 80% of targets, versus 75% for Rosetta.

Conclusions:

• PEP-FOLD2 demonstrates improved accuracy and efficiency in peptide structure prediction.
• The method shows promise for large-scale *in silico* studies, advancing the field of mini-protein structure prediction.
• PEP-FOLD2's performance suggests its maturity as a valuable tool for computational structural biology.