The identification of immunogenic epitopes (such as fragments of proteins, in particular
peptides, that can trigger an immune response) is a fundamental need for immune-based
therapies. A computational tool that could detect such immunogenic epitopes would have vast
potential applications in biomedicine ranging, from vaccine ...»»»»
The identification of immunogenic epitopes (such as fragments of proteins, in particular
peptides, that can trigger an immune response) is a fundamental need for immune-based
therapies. A computational tool that could detect such immunogenic epitopes would have vast
potential applications in biomedicine ranging, from vaccine design against viruses or bacteria
to therapeutic vaccination of cancer patients. While there are several methods that predict
whether a peptide will be shown to the immune system via the HLA proteins of a patient, most
of them cannot predict whether such presentation will indeed trigger an immune response.
The aim of this project is to build an immunogenicity predictor that discriminates immunogenic
from non-immunogenic epitopes. After a careful study of the drivers of antigen processing and
presentation on HLA class I molecules and an assessment of the physicochemical factors
influencing epitope recognition by T-cell receptors (TCRs), we have used a selection of
publicly available tools and in-house developed algorithms to identify the most relevant
features that determine epitope immunogenicity. We then used these features to build an
immunogenicity predictor (PredIG) modelled by logistic regression against immunogenically
validated epitopes by the ImmunoEpitope DataBase (IEDB). Overall, our immunogenicity
predictor shows a better performance in identifying immunogenic epitopes than other stateof-
the-art metrics.^^^^
