This application discloses a method,
system, device, and storage medium for predicting the
immunogenicity of neoantigens. The method includes: acquiring tumor
mutation data and synthesizing candidate peptides; constructing a
training set with HLA
typing and immune tags; extracting and fusing multi-
omics feature vectors covering dimensions such as binding affinity, complex stability, presentation probability, expression level,
mutation frequency, and clonal abundance; training the model using random balanced forest and XGBoost
ensemble learning to obtain an
immunogenicity prediction model; outputting
immunogenicity probability scores for the peptides to be predicted and prioritizing them accordingly to generate a neoantigen
list. This application significantly improves the accuracy and screening efficiency of neoantigen immunogenicity prediction by fusing multi-
omics features and integrating the immunogenicity prediction model.