The invention relates to the technical field of biological information, in particular to an immunoreaction evaluation method based on tumor neoantigen activity sorting. The method comprises the following steps: obtaining
genome data of tumor and normal tissues through whole
exon sequencing, extracting multi-dimensional features of candidate somatic
mutation, and screening by combining a
Gaussian mixture model and a Transform model to obtain a high-confidence
mutation genome set; predicting the HLA
genotype of a patient based on
sequencing data, translating and mutating into
a peptide fragment, predicting the binding affinity of the
peptide fragment and an MHC molecule by using an XGBoost model, and calculating a new
antigen activity
score sequence in combination with various parameters; and finally, synthesizing a new
antigen peptide fragment according to a sorting result, carrying out in-vitro co-culture to detect an IFN-gamma
secretion result, and dynamically optimizing a characteristic combination coefficient through a PPO
algorithm to realize intelligent iterative updating, so that the accuracy of new
antigen screening and the immunoreaction prediction capability are remarkably improved.