PepFold vs PEP-FOLD — Two Different Tools

Side-by-Side Comparison

What PepFold (pepfold.com) Does

PepFold is a pharmacogenomic variant-to-synthesis platform built for researchers, bioinformaticians, and pharmaceutical teams who start with genetic data and need actionable peptide candidates. The platform automates a process that traditionally requires weeks of manual work across multiple disconnected tools and databases.

You provide rsIDs (the unique identifiers for single nucleotide polymorphisms from genotyping or whole genome sequencing results), and PepFold runs a fully automated pipeline:

• Variant annotation via ClinVar: Each rsID is looked up in NCBI ClinVar to determine clinical significance (pathogenic, likely pathogenic, drug-response, etc.), the associated gene, and the review status of the evidence.
• Protein target mapping via UniProt: The affected gene is mapped to its protein product. PepFold retrieves the full protein sequence, functional domains, and binding regions to identify the most promising interaction sites for peptide candidates.
• AI-driven peptide generation: Using NVIDIA BioNeMo Evo 2, a 40-billion parameter genomic foundation model, PepFold generates multiple peptide candidate sequences designed to interact with the target protein. When the Evo 2 API is unavailable, a proprietary rational design engine produces candidates using biochemical complementarity principles.
• 3D structure prediction via ESMFold:Each candidate peptide is folded into a predicted 3D structure using Meta's ESMFold, with per-residue pLDDT confidence scores. These structures are rendered as interactive 3D viewers in the final report.
• 10-dimensional scoring: Candidates are evaluated across binding affinity, structural confidence, clinical relevance, protease stability, ADMET properties, cell permeability, aggregation risk, sequence novelty, half-life estimation, and target selectivity. This multi-dimensional approach ensures that top-ranked candidates are not just theoretically active but practically viable.
• Fmoc-SPPS synthesis protocols: Each top candidate receives a complete solid-phase peptide synthesis protocol, including resin selection, coupling sequence, cleavage conditions, purification parameters, and quality control specifications. These are starting-point templates requiring laboratory validation.

The entire pipeline runs in under two minutes and produces downloadable HTML and PDF reports. PepFold is available as a web application at pepfold.com and as a programmatic API for integration into existing research workflows.

What PEP-FOLD (Université Paris) Does

PEP-FOLD is a de novo peptide structure prediction server developed by the Ressource Parisienne en Bioinformatique Structurale (RPBS) at Université Paris Cité (formerly Université Paris Diderot). It has been an important tool in the structural biology community since its initial publication in 2009, with subsequent versions (PEP-FOLD2 in 2012, PEP-FOLD3 in 2016, and PEP-FOLD4 in 2024) improving accuracy and expanding supported peptide lengths.

PEP-FOLD takes an amino acid sequence as input (typically 5 to 50 residues for PEP-FOLD3, and up to 300 for PEP-FOLD4) and predicts its three-dimensional fold. The method uses a Hidden Markov Model based on a structural alphabet of 27 canonical backbone conformations. This structural alphabet representation is coupled with a coarse-grained molecular simulation to generate and rank candidate 3D conformations.

The tool is free for academic use and hosted as a web server. It does not generate peptide candidates, does not connect to clinical or genomic databases, and does not produce synthesis protocols. Its scope is limited to answering one question: given this amino acid sequence, what does the peptide look like in 3D?

PEP-FOLD has been cited in thousands of research papers and is widely used in the computational biology community. The team behind it, led by Pierre Tufféry and colleagues, has made significant contributions to protein and peptide structure prediction.

When to Use PepFold (pepfold.com)

Choose PepFold when your starting point is genetic data and your goal is therapeutic peptide candidates. Specifically:

• Pharmacogenomic research: You have rsIDs from genotyping panels (23andMe, clinical sequencing, GWAS results) and want to explore peptide-based therapeutic avenues for the identified variants.
• Personalized medicine: You are working on patient-specific treatment strategies and need peptide candidates tailored to specific genetic profiles. PepFold produces candidates informed by both the clinical significance of the variant and the structural properties of the target protein.
• SNP-to-peptide pipelines: You need an automated system that goes from a list of variant identifiers to ranked, scored, structurally predicted peptide candidates with synthesis instructions, without manually querying ClinVar, UniProt, and structure prediction servers separately.
• Drug discovery exploration: You are in the early stages of evaluating peptide therapeutic hypotheses for specific genetic targets and want a rapid computational screen before committing to wet-lab resources.
• Synthesis planning: You need not only candidate sequences but also actionable Fmoc-SPPS protocols that your chemistry team can use as starting points for experimental validation.

PepFold is particularly valuable when you are working with pharmacogenomically relevant variants (drug-response, pathogenic, or likely pathogenic annotations in ClinVar) and need a rapid, reproducible computational pipeline that integrates variant annotation, protein biology, peptide generation, structural prediction, and multi-dimensional scoring in a single automated run.

When to Use PEP-FOLD (Academic)

Choose PEP-FOLD when your starting point is a known peptide sequence and your goal is its predicted 3D structure. Specifically:

• Structure prediction: You have a peptide sequence (either designed manually, obtained from a phage display experiment, or derived from a protein fragment) and need to visualize its predicted fold.
• Conformational analysis: You want to explore multiple potential conformations of a short peptide to understand its structural flexibility and preferred backbone geometry.
• Docking preparation: You need a 3D model of a peptide as input for molecular docking simulations or molecular dynamics studies.
• Academic research: You are in an academic setting, need a free and well-cited structure prediction tool, and your work does not require the upstream genetic analysis or downstream synthesis planning that PepFold provides.

PEP-FOLD is the right choice when you already have your peptide and just need to know what it looks like. It excels at short peptide structure prediction and has been validated extensively in the literature.

Can They Work Together?

Yes. PepFold and PEP-FOLD are complementary tools that address different stages of a peptide research workflow. Using them together can strengthen your results.

Cross-validation of structural predictions

PepFold uses ESMFold (Meta's protein structure prediction model) to generate 3D structures for each candidate. PEP-FOLD uses an entirely different methodology (HMM structural alphabet + coarse-grained simulation). When both tools predict a similar fold for the same peptide sequence, confidence in the structural prediction increases significantly. Conversely, disagreements between the two methods flag candidates that may need additional investigation or experimental validation.

A combined workflow

This combined approach leverages PepFold's end-to-end pharmacogenomic pipeline for candidate generation and scoring, while using PEP-FOLD as an independent structural validation checkpoint. For researchers publishing results, demonstrating structural consensus across multiple prediction methods strengthens the evidence for their candidates.

The Key Difference: Upstream vs. Downstream

The fundamental distinction between PepFold and PEP-FOLD comes down to where each tool sits in the research workflow.

PepFold operates upstream. It starts before you have a peptide. You provide genetic data (variants, SNPs) and the platform generates, evaluates, ranks, and prepares peptide candidates for you. It answers the question: “Given this genetic variant, what therapeutic peptides should I investigate?”

PEP-FOLD operates downstream. It starts after you already have a peptide sequence. You provide the amino acid sequence and it predicts its 3D structure. It answers the question: “Given this peptide sequence, what does it look like?”

Neither tool replaces the other. They serve different needs at different stages. A researcher working on pharmacogenomic peptide therapeutics might use PepFold to generate candidates and PEP-FOLD to cross-validate structures, while a structural biologist studying peptide conformations might only need PEP-FOLD.

Technical Architecture Comparison

For researchers interested in the technical details, here is how the two tools differ architecturally.

PepFold (pepfold.com)

• Data integration layer: Real-time queries to NCBI ClinVar and UniProt REST APIs for variant annotation and protein target mapping.
• Generative layer: NVIDIA BioNeMo Evo 2 (40B parameters) for AI-driven peptide generation, with a proprietary rational design engine as deterministic fallback.
• Structure prediction layer: ESMFold (Meta) for 3D fold prediction with per-residue pLDDT confidence scores.
• Scoring layer: Proprietary 10-dimensional scoring system covering binding affinity, structural confidence, clinical relevance, protease stability, ADMET, permeability, aggregation, novelty, half-life, and selectivity.
• Synthesis layer:Automated Fmoc-SPPS protocol generation tailored to each candidate's sequence properties.
• API: RESTful API for programmatic access and integration into existing bioinformatics pipelines.

PEP-FOLD (Université Paris)

• Structural alphabet: 27 canonical backbone conformations derived from analysis of protein structures, encoded as a Hidden Markov Model.
• Sampling: Coarse-grained molecular simulation (OPEP force field) to explore conformational space.
• Ranking: sOPEP energy scoring function to rank generated conformations.
• Output: Top N 3D models in PDB format.
• Peptide length: Up to 36 residues (PEP-FOLD3), up to 300 residues (PEP-FOLD4 with different methodology).

Common Search Confusions

If you arrived at this page from a search engine, you were likely looking for one of these:

Frequently Asked Questions

Summary

PepFold and PEP-FOLD are two different tools that happen to share a similar name. PepFold (pepfold.com) is a commercial pharmacogenomic platform that takes genetic variants as input and produces ranked, scored, structurally predicted peptide candidates with complete synthesis protocols. PEP-FOLD is a free academic tool that takes peptide sequences as input and predicts their 3D structure.

Both tools are legitimate and valuable in their respective domains. If you are working at the intersection of pharmacogenomics and peptide therapeutics, PepFold provides the end-to-end automated pipeline from variant to synthesis. If you are working on peptide structural biology, PEP-FOLD provides high-quality structure predictions. For the most thorough research, combining both tools leverages complementary prediction methodologies and increases overall confidence in your results.