Vaccine Epitope Selection via HLA Genotype Segmentation

The patent segments the population into distinct HLA genotype groups and creates separate epitope selections for each segment. Instead of using a single supertype-based selection that treats all populations uniformly, the method divides the target population into genetically heterogeneous subgroups and optimizes epitope coverage for each segment independently, thereby resolving the contradiction between simplicity and coverage.

The patent applies local quality by tailoring the epitope selection to match the specific HLA background of each population segment. Rather than applying a uniform supertype-based approach globally, the method adjusts the epitope set locally for each HLA genotype group, ensuring that each segment receives an optimized selection matched to its genetic characteristics.

The patent segments the population into discrete HLA genotype groups and performs separate epitope optimizations for each segment. This segmentation allows the computational method to address individual genetic variations explicitly, ensuring that each citizen group receives an epitope set tailored to their specific HLA background rather than relying on average responses.

The patent changes the optimization parameter from maximizing average response likelihood to maximizing minimum response likelihood across all HLA genotype segments. This parameter transformation shifts the objective function to ensure that even the least-covered genetic group achieves adequate protection, thereby guaranteeing individual citizen protection while maintaining computational tractability.

The patent performs preliminary computational filtering of peptide candidates based on predicted binding affinity to multiple HLA alleles before conducting experimental screening. By pre-screening in silico and selecting only the most promising candidates for wet-lab validation, the method maintains high epitope identification accuracy while dramatically reducing the time and resources required for experimental screening of large peptide sets.

The patent uses computational models and algorithms to create virtual representations of peptide-MHC interactions, allowing in silico testing of numerous peptide candidates before physical experimentation. This copying approach enables accurate prediction of epitope binding without the need for extensive experimental screening of every candidate, thereby reducing time and cost while maintaining identification accuracy.