In silico Immunogenicity Assessment for Sequences Containing Unnatural Amino Acids: A Method Using Existing in silico...

Area of Science:

• Immunoinformatics
• Computational Biology
• Drug Development

Background:

• T cell epitope prediction is crucial for assessing drug immunogenicity.
• Standard immunoinformatics tools struggle to accurately estimate HLA binding for peptides containing unnatural amino acids (UAAs).
• UAAs are increasingly used in peptide therapeutics and can be present as impurities.

Purpose of the Study:

• To develop an in silico method for predicting the impact of UAAs on peptide-HLA binding.
• To enhance the assessment of immunogenic potential for peptide drug candidates and impurities.
• To reduce the cost and improve the efficiency of immunogenicity evaluation.

Main Methods:

• Developed a computational method to model UAA effects on peptide-HLA binding.
• Utilized in silico assessment for risk-based selection of peptide candidates.
• Employed in vitro HLA binding studies to estimate UAA binding potentials and refine computational predictions.

Main Results:

• Demonstrated in silico immunogenicity prediction for common impurities in teriparatide and semaglutide.
• Showcased the use of in vitro studies to correct in silico binding estimates for UAAs.
• Provided an example using D-amino acids in the PADRE peptide to illustrate UAA impact.

Conclusions:

• The developed in silico method accurately models UAA influence on peptide-HLA binding and immunogenicity.
• Integration of in vitro and in silico approaches provides a robust strategy for evaluating peptide immunogenicity.
• Future predictive models can be established with increased HLA binding data for UAAs, enabling direct estimation.