Advanced Cyclic Peptide Structure Prediction with HighFold Now Available on Vecura

What is HighFold?

HighFold is a specialized computational framework designed for the accurate 3D structure prediction of cyclic peptides, including both head-to-tail monomers and complex peptide-protein interactions. By extending the capabilities of AlphaFold2 and ColabFold, it introduces a Cyclic Position Offset Encoding Matrix (CycPOEM) to explicitly model the topological constraints of cyclic ring structures. This framework overcomes the fundamental limitation of standard protein folding models, which typically treat all input sequences as linear chains.

It helps researchers model the preferred geometry of macrocycles, incorporate disulfide bridge constraints, and simulate binding modes between cyclic peptides and protein targets. It is especially useful for drug discovery, antimicrobial research, and the design of peptide-based therapeutics.

What can users do with HighFold on Vecura?

With HighFold on Vecura, users can:

• Predict the 3D structures of cyclic peptide monomers with head-to-tail cyclization.
• Model the docking of cyclic peptides against target proteins using AlphaFold2-multimer.
• Apply user-defined constraints for intra-peptide disulfide bridges to enhance structural accuracy.
• Generate high-confidence PDB files, complete with pLDDT, pTM, and ipTM scoring metrics to evaluate binding interfaces and structural reliability.

What the output means

The output provides a 3D structural model in PDB format, accompanied by critical confidence metrics such as mean pLDDT and pTM scores. For complex predictions, an interface-specific ipTM score is generated to assess the validity of the peptide-protein binding interface.

This output should be used to support scientific decision-making, such as screening candidate designs or rationalizing structure-activity relationships (SAR). It does not replace experimental validation, such as X-ray crystallography or NMR spectroscopy.

Why this matters

Cyclic peptides are a vital class of molecules in pharmacology, prized for their metabolic stability and ability to target "undruggable" protein interfaces. Despite their importance, conventional structure prediction tools often fail to correctly model the ring closure and the geometric constraints unique to these scaffolds.

HighFold bridges this gap by integrating geometric constraints directly into the Evoformer stack of AlphaFold2. By providing a streamlined, accessible workflow for predicting these complex geometries, HighFold enables researchers to prioritize promising peptide scaffolds more efficiently, ultimately accelerating the drug development pipeline.

• Developed by: Hongliang Duan et al.
• Source: Official GitHub Repository
• Reference: HighFold Documentation