netmhcpanwrapper 0.1.0

Python wrapper for NetMHCpan-4.2 binding predictions

Python NetMHCpan Wrapper

Seamless integration of NetMHCpan binding predictions into pandas-based bioinformatics workflows.

Description

This package provides a user-friendly Python interface to NetMHCpan-4.2 (standalone version).
It handles peptide-MHC class I binding affinity predictions while solving problems of inconsistent HLA nomenclature and missing HLA information.

The wrapper supports two main prediction regimes: single allele and pan/supergroup prediction, with built-in parallel processing for speed.

Goals

• Make NetMHCpan easy to use inside Python/pandas scripts and notebooks
• Automatically fix inconsistent HLA allele names (e.g. HLA-A*01:01, A0101, A01)
• Enable pan-prediction when only broad allele families are known

Supported Regimes

1. Single Mode (single)

• Requires an HLA column
• Automatically standardizes allele names using a mapping file
• Predicts binding for the correctly matched allele

Best for: Standard epitope-HLA tables from IEDB, VDJdb, or custom datasets.

2. Pan / Supergroup Mode (pan)

• Works with or without a helper (supergroup) column
• Supports predictions for Homo sapiens ('hs'), Mus musculus ('mmu) or pan-prediction without species specification
• If a valid family is given (e.g. HLA-A01, HLA-B15, HLA-E01), searches only inside that family
• If no family or unknown family → performs a smart two-round search:
  1. First round: finds the best supergroup using representative alleles
  2. Second round: finds the best allele inside the winning supergroup
• Handles HLA-free epitope lists

Best for: Large epitope sets with partial or no HLA information.

Directory Structure

mhc_netmhcpan_wrapper/
├── src/
│   ├── __init__.py
│   ├── config.py
│   ├── run_netmhcpan.py          # Main high-level function
│   ├── single_predictor.py
|   ├── mhc_supergroups_hs.txt
|   ├── mhc_supergroups_mmu.txt
│   └── pan_predictor.py
├── datasets/
│   ├── allele_nomenclature_mapping.txt
│   └── mhc_supergroups.txt
├── tmp/                          # Auto-created for temporary files
├── testing.ipynb                 # Usage examples
├── Netmchpan_wrapper_development_notebook.ipynb
├── README.md
└── ...

---

Quick Start Tutorial

1. Configuration

Edit src/config.py and set the number of cores:

from pathlib import Path

TMPDIR = Path('../tmp')
PATH_TO_SUPERGROUPS = Path('../datasets/mhc_supergroups.txt')
PATH_TO_MAPPING = Path('../datasets/allele_nomenclature_mapping.txt')
N_CORES = 8

2. Basic Usage

import pandas as pd
from pathlib import Path
from src.run_netmhcpan import run_netmhcpan

PATH_TO_NETMHCPAN = Path("/path/to/netMHCpan-4.2bstatic.Linux/netMHCpan-4.2")

# === Single mode ===
df = pd.read_csv("your_data.tsv", sep="\t")

result = run_netmhcpan(
    prediction_mode='single',
    path_to_netmhcpan=str(PATH_TO_NETMHCPAN),
    df=df,
    epitope_colname='antigen.epitope',
    hla_column='hla'
)

# === Pan mode ===
pan_df = pd.DataFrame({
    'epitope': ['LIDGIFLRY', 'VMADRTRHL', 'ANADLEVKI'],
    'family': [None, 'HLA-E01', 'HLA-C06']
})

result_pan = run_netmhcpan(
    prediction_mode='pan',
    path_to_netmhcpan=str(PATH_TO_NETMHCPAN),
    df=pan_df,
    epitope_colname='epitope',
    supergroup_column='family'   # can be None for pure pan search
    species = 'None' #can be hs or mmu for species-restricted pan-prediction
)

---

Using Classes Directly (Advanced)

from src.single_predictor import SinglePredictor

single = SinglePredictor(
    path_to_netmhcpan=PATH_TO_NETMHCPAN,
    path_to_mapping=Path("datasets/allele_nomenclature_mapping.txt"),
    tmpdir=Path("tmp"),
    n_cores=8
)

df = single.match_hla(df, 'hla')
df = single.predict_affinity_dataframe(df, 'epitope', 'matched_hla')

---

Important Notes

• You must have NetMHCpan-4.2 installed and working from the command line (netMHCpan -h)
• The wrapper calls the binary via subprocess
• Predictions are run in BA mode (-BA)
• For very large datasets, adjust N_CORES according to your available CPU and memory

---

License

This wrapper is available under GNU General Public License.

NetMHCpan is distributed by DTU Health Tech (https://services.healthtech.dtu.dk/services/NetMHCpan-4.1/) — please respect their licensing terms.