nf-core pipeline · epitope, epitope-prediction, mhc-binding-prediction

nf-core/epitopeprediction

A Nextflow pipeline that predicts peptide binding to MHC class I and II molecules from variants, proteins, or peptides.

Run it with the research agent Documentation Source code

What nf-core/epitopeprediction is

nf-core/epitopeprediction is an analysis pipeline designed to predict peptide binding affinity to major histocompatibility complex (MHC) molecules. MHC molecules play a key role in the immune system by presenting short peptide fragments on the cell surface for recognition by T cells. Identifying which peptides successfully bind given MHC alleles is crucial for neoepitope discovery, immunogenicity assessments, and vaccine design.

Originally written by Christopher Mohr and Alexander Peltzer, with contributions from Sabrina Krakau, Leon Kuchenbecker, Marissa Dubbelaar, Gisela Gabernet, Jonas Scheid, and Alina Bauer, the workflow standardises epitope binding predictions. It eliminates manual multi-step processing by orchestrating peptide generation, running established prediction software, and combining output scores into unified tables.

The pipeline accommodates diverse research entry points, including somatic mutation lists, full protein sequences, and raw peptide sets. It processes these inputs across user-specified MHC alleles and collates results into sample-level summaries alongside an interactive MultiQC evaluation report.

What the pipeline does

The pipeline executes in four main stages. First, candidate peptides are produced from the provided input: somatic variant call format (VCF) files are filtered and normalised using bcftools, annotated with Ensembl VEP, and translated into mutant peptides using pVACtools; protein sequences in FASTA format are cleaved into peptides of user-defined lengths; and peptide lists in TSV format are processed directly.

Next, the peptides are evaluated for MHC binding against the alleles defined in the sample sheet. Predictions can be made using MHCflurry, MHCnuggets (the default tool), MHCnuggetsII, NetMHCpan, NetMHCIIpan, MixMHCpred, or MixMHC2pred. Finally, predictions across all specified tools are aggregated into a single tabular output per sample, and overall binding statistics are compiled into a MultiQC report.

Key facts

Pipeline engine: Nextflow (DSL2)
Default prediction tool: MHCnuggets
Supported prediction tools: MHCflurry, MHCnuggets, MHCnuggetsII, NetMHCpan, NetMHCIIpan, MixMHCpred, MixMHC2pred
Supported inputs: VCF (variants), FASTA (proteins), TSV (peptides)
Primary outputs: Aggregated prediction tables per sample, MultiQC report
Container support: Docker, Singularity

What you can use it for

Neoepitope discovery from somatic variants: Input somatic variant calls (VCF) alongside a VEP cache and reference FASTA to translate tumour mutations into novel mutant peptides and predict binding to patient-specific MHC alleles.
Antigenic region scanning in protein sequences: Supply full-length protein FASTA files to digitally digest proteins into overlapping peptide fragments and scan for MHC binding regions.
Immunopeptidomics binding validation: Provide mass spectrometry-derived peptide lists in TSV format to predict and annotate MHC class I or class II binding affinity across corresponding host alleles.

Run it here

The research agent runs nf-core/epitopeprediction with Nextflow on its own machine — up to 32 cores, 256 GiB of memory and 24 hours per run, which covers the test profile and real single-sample runs. It builds the samplesheet, checks it against the pipeline's schema and reads the results back. When a run is too big for one machine, it gives you the exact command for your own cluster instead.

Ask the research agent

Strengths and limits

Good at

Accepts three distinct biological data types: somatic variants, protein sequences, and explicit peptide lists.
Integrates multiple widely used MHC class I and II prediction algorithms within a single harmonised interface.
Consolidates predictions from multiple algorithms into a unified, sample-specific output table alongside a MultiQC summary.
Implements Nextflow DSL2 with isolated process containers, ensuring modular execution and reproducibility across computing environments.

Watch out for

Processing somatic variant inputs requires external reference data, specifically an Ensembl VEP cache and a genome FASTA file.
Prediction tools such as NetMHCpan, NetMHCIIpan, MixMHCpred, and MixMHC2pred are not open-source by default and require separate user compliance with their respective upstream licenses.

Questions

Can I run nf-core/epitopeprediction without my own cluster?

What licence is nf-core/epitopeprediction under?

MIT, like every nf-core pipeline.

What input formats are supported by nf-core/epitopeprediction?

The pipeline accepts somatic variant files in VCF format, protein sequences in FASTA format, or peptide lists in TSV format, configured through a standard CSV sample sheet.

Which MHC prediction software does the pipeline use by default?

The pipeline runs MHCnuggets by default, but users can configure it to use alternative predictors including MHCflurry, MHCnuggetsII, NetMHCpan, NetMHCIIpan, MixMHCpred, and MixMHC2pred.

Can the pipeline predict both MHC class I and class II binding?

Yes, it supports both MHC class I and class II predictions when supplied with the relevant alleles and compatible tools, such as NetMHCpan/NetMHCIIpan, MixMHCpred/MixMHC2pred, or MHCnuggets/MHCnuggetsII.

What extra reference files are required when starting from VCF files?

Running the pipeline on somatic VCF files requires a reference genome FASTA file and an Ensembl Variant Effect Predictor (VEP) cache to perform variant annotation and mutant peptide generation.

How to cite it

doi: 10.5281/zenodo.3564666

Written from nf-core/epitopeprediction's own README. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.