Accurate prediction of immunogenic T-cell epitopes from epitope sequences using the genetic algorithm-based ensemble...

Area of Science:

• Immunology
• Bioinformatics
• Vaccine Design

Background:

• T-cell epitopes are vital for T-cell immune responses and epitope-based vaccine development.
• Immunogenicity prediction is significant for effective vaccine design and understanding immune system functions.
• Accurate identification of immunogenic T-cell epitopes is a key challenge in immunology and vaccinology.

Purpose of the Study:

• To develop a computational method for differentiating immunogenic from non-immunogenic T-cell epitopes based on primary structures.
• To explore and analyze sequence-derived features relevant to epitope immunogenicity.
• To optimize feature selection and build a high-accuracy ensemble model for immunogenicity prediction.

Main Methods:

• Investigated various sequence-derived features to identify their relationship with T-cell epitope immunogenicity.
• Employed a genetic algorithm (GA)-based ensemble method for optimal feature subset selection and model development.
• Utilized cross-validation with AUC (area under ROC curve) as the primary metric for GA optimization.

Main Results:

• The GA-based ensemble method achieved superior performance on benchmark datasets (IMMA2 and PAAQD) compared to state-of-the-art methods.
• Achieved AUC scores of 0.846 on the IMMA2 dataset and 0.829 on the PAAQD dataset.
• Statistical analysis confirmed the significant performance improvements of the proposed GA-based ensemble method.

Conclusions:

• The proposed GA-based ensemble method serves as a promising computational tool for predicting immunogenic T-cell epitopes.
• The developed method offers enhanced accuracy for identifying epitopes critical for vaccine design.
• Source codes and datasets are provided for reproducibility and further research.