Summary: 

The National Cancer Institute (NCI) seeks licensees for a machine learning algorithm that scores epitopes for likelihood of reactivity in order to create personalized effective immunotherapy.

Description of Technology: 

Success in immunotherapy is often attributable to the reactivity of patient T-cells to specific mutated peptide(s) found in the patient’s tumor known as neoepitopes. In the development of patient-specific immunotherapies, there is no consistent standard for prioritizing such neoepitopes. Current models arrive at a ranked list of potential candidates by removing epitopes based on pre-determined criteria which might lead to the elimination of known reactive neoepitopes. 

Identification, prioritization and targeting of patient neoepitopes are crucial for developing effective, personalized treatments. Ranking or prioritizing neoepitopes is especially important when trying to construct a cancer vaccine that will elicit a therapeutically beneficialn immune response. Accordingly, scientists at the National Cancer Institute (NCI) have created a novel approach to identify and prioritize patient neoantigens. This model uses a training dataset of known neoantigens from patient screening and determines features of importance to epitope recognition using both reactive and non-reactive epitopes. The machine learning algorithm scores epitopes for their likelihood of reactivity and provides a stable, reproducible method to prioritize epitopes that can be used anywhere in the world. 
 
The National Cancer Institute (NCI) seeks licensees for this machine learning algorithm that scores epitopes for likelihood of reactivity in order to create personalized effective immunotherapy.

Potential Commercial Applications: 

•    Oncology
•    Prioritization of neoantigens for the development of effective personalized therapies
o    Cancer vaccines
o    TIL and T-cell receptor therapies
•    Research use

Competitive Advantages:

•    Model is trained using a dataset of verified neoantigens from patient tumor data 
•    Model is unbiased because it does not use prior assumptions about what features a neoepitope should have
•    Uses two models (MMP and NMER model) aswhich is a more reproducible approach than using a single model
•    Particularly useful for prioritizing epitopes for patients with large numbers of mutations