Deep learning using tumor HLA peptide mass spectrometry datasets improves neoantigen identification

Area of Science:

• Oncology
• Immunology
• Bioinformatics

Background:

• Neoantigens are key targets for antitumor T-cell responses.
• Current neoantigen identification methods are invasive, costly, or limited in scope.
• Developing effective neoantigen-targeted cancer immunotherapies is a major goal.

Purpose of the Study:

• To develop a computational model for accurate neoantigen prediction.
• To overcome limitations of existing neoantigen identification techniques.
• To facilitate the development of novel cancer immunotherapies.

Main Methods:

• Applied deep learning to a dataset of 74 patients' tumor genomic and HLA peptide data.
• Developed a computational model named EDGE for antigen presentation prediction.
• Validated the model's performance on predicting neoantigens.

Main Results:

• The EDGE model significantly improved the positive predictive value of HLA antigen prediction by up to ninefold.
• Enabled neoantigen identification using routine clinical specimens and minimal synthetic peptides.
• Demonstrated applicability across most common human leukocyte antigen (HLA) alleles.

Conclusions:

• EDGE offers a more efficient and accessible approach for neoantigen prediction.
• Facilitates the development of neoantigen-targeted immunotherapies for cancer patients.
• Represents a significant advancement in computational cancer immunology.