Step 1: Epitope identify
VLPIM uses NetMHCIIpan for CD4+ T-cell epitope identification.
Upload files for epitope identification
Upload NetMHCIIpan output file and enter VLP sequence for analysis
Reduce mode: prioritize higher binding count score (favoring immunogenicity reduction).
Enhance mode: prioritize lower binding count score (favoring immunogenicity enhancement).
Identified epitopes ({{ advancedEpitopes.length }})
| Sequence | Length | Start | End | Core | Allele | IC50 (nM) | %Rank_EL | Binding Count |
|---|---|---|---|---|---|---|---|---|
{{ row.sequence }} |
{{ (row.sequence || '').length }} | {{ row.start }} | {{ row.end }} | {{ row.core }} |
{{ row.allele || '-' }} | {{ row.ic50 || '-' }} | {{ row.rank_el || row.rank || '-' }} | {{ row.number_of_strong_binding || 0 }} |
Identified epitopes
| Peptide | Allele | IC50 (nM) | Rank (%) | Position |
|---|---|---|---|---|
{{ row.Peptide || row.peptide }} |
{{ row.Allele || row.allele }} | {{ row['BA_IC50'] || row.IC50 || row.ic50 || '-' }} | {{ row.Rank || row.rank || '-' }} | {{ row.Pos || row.pos || '-' }} |
Key features:
- Epitope prediction for full VLP sequence using NetMHCIIpan
- Multi-allele HLA-DRB1 binding evaluation
- Epitope ranking based on binding performance (sum of strong and weak binders, Binding count). A higher score indicates epitopes more suitable for immunogenicity reduction, whereas a lower score suggests epitopes favorable for immunogenicity enhancement.
- Customizable peptide length and number of epitopes (default: 15 amino acids and 10 epitopes)
Step 2: Mutation pool generation
Generate a mutation pool for downstream immunogenicity optimization using ProteinMPNN.
ProteinMPNN setup and execution parameters
Refer to the tool setup for recommended parameters, input formats, and example commands.
Key points:
- Batch generation of variants for pool generation
- ProteinMPNN parameters: sampling temperature (0.1–0.5), sequences per target (50–200), fixed-backbone mode, residue masking for conserved regions
- Inputs/outputs: input PDB/CIF (with chain selection); output FASTA/CSV sequence pool for Step 3
Step 3: Immunogenicity and structure prediction
Evaluate the mutation pool with NetMHCIIpan (immunogenicity) and AlphaFold3 (structure prediction), then proceed to Step 4.
Configuration of NetMHCIIpan and AlphaFold3
See tool setup for command templates, inputs/outputs, and post-processing recommendations.
Key points:
- Feed NetMHCIIpan outputs (immunogenicity) outputs and AlphaFold3 outputs(structure) to Step 4.1 and 4.2 for detailed analysis
- Keep inputs/outputs consistent for traceability and reproducibility
- NetMHCIIpan parameters: provide HLA-DRB1 allele list; interpret %Rank thresholds (strong ≤ 1.0, weak ≤ 5.0); export wide-table output
- AlphaFold3 parameters: set recycles/seeds; optionally use a reference for Step 4.2; retain predicted PDB/CIF for RMSD/TM-score
Step 4.1: Immunogenicity analysis
Upload files for immunogenicity analysis
Upload NetMHCIIpan output file，then choose mode and compute.
Reduce mode: favoring decreased immunogenicity.
Peptide immunogenicity scores
| Rank | ID | Peptide | Alleles | Best IC50 (nM) | Sim score |
|---|---|---|---|---|---|
| {{ row.Rank || (idx + 1) }} | {{ row.ID || '-' }} |
{{ row.Peptide }} |
{{ row.Allele_Count || row.Alleles || '-' }} | {{ row.Best_IC50 || row['Best_IC50(nM)'] || '-' }} |
Key features:
- Enhance and reduce modes to prioritize sequences or peptides
- In both enhance and reduce modes, a lower Sim score corresponds to a greater change in immunogenicity, providing a standardized metric for prioritizing candidate mutations
- Sorting by Sim score, best IC₅₀, or allele count
- Export combined results to CSV with one click
Step 4.2: Structural superposition
Upload AlphaFold3 predicted structures and compare with reference structures using RMSD analysis.
RMSD structure comparison
Upload reference structure (original PDB) and predicted structure (AF3 result PDB) to calculate RMSD.
Structure alignment result
TM-score: {{ rmsdResult.tmScore.toFixed(4) }} Similar fold (TM-score ≥ 0.5) Different fold (TM-score < 0.5)
RMSD: {{ rmsdResult.rmsd.toFixed(3) }} Å
Aligned residues: {{ rmsdResult.alignedAtoms }} / {{ rmsdResult.totalAtoms || rmsdResult.alignedAtoms }}
Iterations: {{ rmsdResult.cyclesUsed }}
Status: Good structural similarity (RMSD < 2.0 Å) Higher deviation (RMSD ≥ 2.0 Å)
Structural superposition features:
- Kabsch algorithm: Standard RMSD calculation based on optimal rotation matrix
- Iterative pruning: Iteratively remove outliers to improve alignment accuracy
- TM-align: Template Modeling score (0-1 scale) for fold similarity assessment. TM-score ≥ 0.5 indicates similar folds
- Cα atom alignment: Automatic superposition of Cα atoms for accurate comparison
- Quality assessment: RMSD < 2.0 Å or TM-score ≥ 0.5 indicates structural similarity
Documentation
README
Comprehensive guide covering installation, configuration, and usage of VLPIM.
Includes quick start guide, prerequisites, and troubleshooting.
Tool setup
Usage parameters for ProteinMPNN, NetMHCIIpan, AlphaFold3, and Rosetta wrappers are documented on the dedicated Tool setup page.
Contact us
Get help and support from the VLPIM team, Research Center for Medicinal Structural Biology, National Research Center for Translational Medicine at Shanghai, State Key Laboratory of Medical Genomics, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China.
Email: wcf231229@163.com
Issues: GitHub Issues
Citation
If you use VLPIM in your research, please cite:
@software{vlpim,
title={VLPIM: A Comprehensive Tool for Immunogenicity Modulation of Virus-like Particles},
author={Chufan Wang},
year={2025},
url={https://github.com/RuijinHospitalVNAR/Toolboxes}Acknowledgments
- ProteinMPNN - Sequence generation
- NetMHCIIpan - MHC-II binding prediction
- AlphaFold3 - Structure prediction
- Rosetta - Interface analysis
Analytics dashboard
Analytics is optional and disabled by default. Enable anonymous analytics in the footer to load this dashboard.