Skip to content

Antigen Processing and Presentation Scores

OptiMHC integrates three external tools for predicting MHC binding affinity and antigen presentation. These tools provide complementary information about whether a peptide is likely to be naturally presented on the cell surface by MHC molecules.

NetMHCpan (MHC Class I)

Config name: NetMHCpan | Source name: NetMHCpan

NetMHCpan 4.1 predicts peptide binding to MHC Class I molecules using artificial neural networks trained on binding affinity and eluted ligand data.

Output Columns

Column Description
netmhcpan_best_score Binding score for the best allele
netmhcpan_best_affinity Predicted binding affinity (nM) for the best allele
netmhcpan_best_percentile_rank Percentile rank for the best allele

Computation

  1. Preprocess peptide sequences: strip flanking amino acids and remove modifications.
  2. Filter peptides to length 8–30 (the supported range for NetMHCpan).
  3. Predict binding for all peptide-allele combinations using NetMHCpan 4.1.
  4. Best allele selection: for each peptide, select the allele \( a^* \) with the minimum percentile rank \( r \):
\[ a^* = \arg\min_{a} \; r(\text{peptide}, a) \]

Missing values (peptides outside the length range) are filled with the column median.

Configuration

featureGenerator:
  - name: NetMHCpan

External installation required

NetMHCpan is a standalone executable that must be downloaded from DTU Health Tech and added to your PATH. See Installation for details.

NetMHCIIpan (MHC Class II)

Config name: NetMHCIIpan | Source name: NetMHCIIpan

NetMHCIIpan 4.3 predicts peptide binding to MHC Class II molecules. It uses binding affinity (BA) mode.

Output Columns

Column Description
netmhciipan_best_score Binding score for the best allele
netmhciipan_best_affinity Predicted binding affinity (nM) for the best allele
netmhciipan_best_percentile_rank Percentile rank for the best allele

Computation

  1. Preprocess peptide sequences: strip flanking amino acids and remove modifications.
  2. Filter peptides to length 9–50 (the supported range for NetMHCIIpan).
  3. Predict binding for all peptide-allele combinations using NetMHCIIpan 4.3 BA.
  4. Best allele selection: for each peptide, select the allele with the minimum percentile rank.

Missing values are filled with the column median.

Configuration

featureGenerator:
  - name: NetMHCIIpan

External installation required

NetMHCIIpan is a standalone executable that must be downloaded from DTU Health Tech and added to your PATH. See Installation for details.

MHCflurry (MHC Class I)

Config name: MHCflurry | Source name: MHCflurry

MHCflurry provides Class I MHC binding affinity predictions along with antigen processing and presentation scores. Unlike NetMHCpan, MHCflurry also models the antigen processing pathway (proteasomal cleavage and TAP transport) in addition to MHC binding.

Output Columns

Column Description
mhcflurry_affinity Predicted binding affinity (nM)
mhcflurry_processing_score Antigen processing score (likelihood of proteasomal cleavage and TAP transport)
mhcflurry_presentation_score Combined presentation score integrating binding and processing
mhcflurry_presentation_percentile Percentile rank of the presentation score

Computation

  1. Preprocess peptide sequences: strip flanking amino acids and remove modifications.
  2. Filter peptides to length 8–15 (the supported range for MHCflurry).
  3. Predict using Class1PresentationPredictor, which returns:
    • Affinity — predicted IC50 binding affinity in nM.
    • Processing score — a score reflecting the likelihood that the peptide undergoes proper antigen processing (proteasomal cleavage, TAP transport).
    • Presentation score — a combined score that integrates binding affinity and processing likelihood.
    • Presentation percentile — the percentile rank of the presentation score relative to a reference distribution.

Missing values (peptides outside the 8–15 length range) are filled with the column median.

Configuration

featureGenerator:
  - name: MHCflurry

No additional parameters are required. The alleles are taken from the top-level allele configuration.

Choosing Between Tools

Tool MHC Class Peptide Length Scores External Setup
NetMHCpan I 8–30 Binding score, affinity, percentile rank Manual download, add to PATH
NetMHCIIpan II 9–50 Binding score, affinity, percentile rank Manual download, add to PATH
MHCflurry I 8–15 Affinity, processing, presentation, percentile None (pip installed)

For MHC Class I analyses, you can use both NetMHCpan and MHCflurry simultaneously — they provide complementary predictions. For MHC Class II, NetMHCIIpan is the only option among these three tools.