PEP-FOLD

PEP-FOLD predicts three-dimensional conformations of peptides to support structural characterization and rational peptide design.

Key Features:

• De Novo Modeling: Predicts 3D conformations for peptides (originally 9–25 amino acids) in aqueous solution using a hidden Markov model-derived structural alphabet and a greedy assembly guided by a modified OPEP coarse-grained force field.
• Simulation and Conformation Selection: Performs multiple simulations (typically 50) and identifies representative conformations based on energy and population metrics.
• Benchmarking Accuracy: Locates low-energy conformations with average deviations around 2.6 Å from full NMR structures for peptides between 9–23 amino acids.
• Expanded Capabilities (PEP-FOLD2 and Beyond): Extends modeling to linear and disulphide-bonded cyclic peptides up to 36 amino acids and supports the definition of disulphide bonds and residue–residue proximities.
• Large-Scale Prediction (PEP-FOLD3): Introduces a computational framework capable of predicting free or biased conformations for linear peptides of 5–50 amino acids and generating native-like conformations for peptides interacting with proteins when the interaction site is known.
• Comparative Performance: Demonstrates superior performance in benchmarks versus PEP-FOLD1 and the Rosetta program for producing near-native or native models of structurally diverse peptides.
• Computation Speed: Returns predictions rapidly (often within minutes), enabling large-scale prediction campaigns.

Scientific Applications:

• Structural Biology: Provides 3D peptide models to support structural analysis and interpretation.
• Chemical Biology: Aids rational design and characterization of peptides with defined conformations.
• Therapeutic Peptide Development: Supports design and assessment of candidate therapeutic peptides by supplying predicted structures.
• Peptide–Protein Interaction Modeling: Generates peptide conformations suitable for modeling peptide–protein interactions when the binding site is specified.
• Large-Scale Screening: Enables high-throughput prediction of peptide conformations for library-scale studies.

Methodology:

Uses a hidden Markov model-derived structural alphabet (SA) of 27 four-residue letters, a greedy assembly procedure guided by a modified OPEP coarse-grained force field, multiple simulations (typically ~50), and selection of representative conformations based on energy and population metrics; PEP-FOLD3 adds a framework for predicting free or biased conformations.