POPISK

Popisk predicts the immunogenicity of HLA-A2-binding peptides to forecast T-cell reactivity and identify sequence positions that influence T-cell receptor recognition.

Key Features:

• Immunogenicity Prediction: Predicts the immunogenic potential of peptides binding HLA-A2 by evaluating their ability to elicit T-cell responses.
• Recognition Position Identification: Identifies key peptide positions (4, 6, 8, and 9) that influence T-cell receptor (TCR) recognition.
• Machine Learning Algorithm: Implements support vector machines (SVM) with weighted degree string kernels for sequence-based classification.
• MHC Restriction Consideration: Classifies peptides according to associated MHC alleles to account for the molecular context of peptide–MHC interactions.
• Predictive Accuracy: Achieves a mean 10-fold cross-validation accuracy of 68% for predicting T-cell reactivity among HLA-A2-binding peptides.
• Mutation Impact Analysis: Predicts changes in immunogenicity caused by point mutations within epitopes, consistent with findings from crystallography.
• Physicochemical Insights: Correlates recognition positions with physicochemical properties and structural features of MHC–peptide–TCR interactions.

Scientific Applications:

• Vaccine Design: Supports design of peptides with enhanced immunogenic profiles for vaccine development.
• Epitope Engineering: Assesses the impact of point mutations on epitope immunogenicity to guide epitope modification and selection.
• Immunology Research: Elucidates molecular determinants of peptide–MHC–TCR interactions and antigen processing and presentation.
• Structural Correlation: Links sequence-based immunogenicity predictions to crystallographic structural observations.

Methodology:

Popisk trains a support vector machine with weighted degree string kernels on a large dataset compiled from major immunology databases.