Exploring diverse approaches for predicting interferon-gamma release: utilizing MHC class II and peptide sequences.

Area of Science:

• Biotechnology
• Immunology
• Computational Biology

Background:

• Therapeutic proteins require immunogenicity assessment for safety and efficacy.
• Current experimental methods for assessing immunogenicity are costly and time-consuming.
• Screening diverse peptide sets across Major Histocompatibility Complex (MHC) alleles is challenging.

Purpose of the Study:

• To develop a computational classification model for predicting interferon-gamma release.
• To utilize peptide sequence and MHC class II (MHC-II) allele pseudo-sequence for prediction.
• To enhance the screening of immunogenic peptides for therapeutic proteins.

Main Methods:

• Utilized a dataset from the Immune Epitope Database, labeled as active or inactive.
• Employed a random forest algorithm with letter-based encoding for classification.
• Evaluated model generalizability using a T-cell proliferation dataset.

Main Results:

• The random forest model with letter-based encoding demonstrated superior performance in predicting interferon-gamma release.
• Feature importance analysis provided insights into model decision-making.
• Virtual single-point mutations enhanced model interpretability.

Conclusions:

• A computational approach can effectively predict T-cell response to therapeutic proteins.
• The developed model offers a more efficient alternative to experimental immunogenicity assays.
• Further research can leverage this model for improved drug development and safety assessment.