A machine-learning platform to predict antibody-stimulating neoantigens from liquid biopsies

During Q1 2022 NOI has begun the process of validating the predictions from the B-cell epitope predictor both internally and in collaboration with Professor Ola Myklebost's lab at the university of Bergen. Internally, validation has been performed using independent test sets for both linear and conformational B-cell epitopes and has also been tested on well characterized immunologically relevant proteins such as the spike protein from SARS-CoV-2. The results look very promising and significantly out-perform other published predictors in benchmark tests. The validation using patient samples from sarcoma patients at the University of Bergen is ongoing. Plasma samples from the sarcoma patients and healthy donor controls have been profiled using peptide and whole protein arrays to identify tumor specific antibody responses. However, to-date the measured antibody responses do not appear to target proteins containing tumor-specific mutations identified using NGS. We now plan to validate the predictions in other cancer types which have a larger mutational burden and may be better suited for this purpose. NOI has also been using the B-cell predictor developed in this project to analyze putative targets from infectious diseases and has recently been awarded prestigious funding from the Coalition for Epidemic Preparedness Foundation (CEPI) worth approximately 5m USD to leverage this the B-cell predictor in combination with the NEC Immune Profiler to develop a blueprint for a pan beta-coronavirus vaccine.

OUTCOMES The project has enabled NOI to: 1. Develop a ML framework for predicting B-cell epitopes in proteins where no prior 3D structure exists, which can be applied to neoantigen prediction and the development of vaccine blueprints against infectious diseases. 2. The results helped NOI to (A) extend the capability of the NIP software to predict non-HLA restricted neoantigens, (B) win an award from CEPI to develop a pan beta-coronavirus vaccine, and (C) encouraged NOI to establish an infectious disease business unit. IMPACTS 1. Enable more cancer patients to benefit from personalized immunotherapies. 2. Enable the development of "liquid biopsy"-based diagnostic methods that can identify cancer patients with early-stage disease helping to save lives and reduce the health and economic societal costs. 3. Facilitate the development of new vaccines to address both emerging and endemic infectious diseases helping reduce the mortality and cost of future pandemics.

There is increasing evidence that neoantigens can be used as generic biomarkers for patient stratification, and represent attractive targets for developing personalized cancer immunotherapies. However, the field has focused on HLA-restricted neoantigens which activate the cellular arm of the immune system (T-cells), and has largely ignored non-HLA-restricted antigens that can activate the humoral arm of the immune system i.e. generate antibodies. Furthermore, all current neoantigen screening and prediction methodologies requires access to tumor biopsy material - which is not always possible to access. The goal of this project is to develop an innovative machine-learning platform named NeoAbScan that can predict non-HLA-restricted neoantigens from next generation sequencing (NGS) data from a patient biopsy, but also from non-invasive liquid biopsies. The outcome will be a tool that can generate a more holistic immune-system-orientated overview of a tumor's neoantigen landscape, which will improve patient stratification, patient monitoring and also broaden the repertoire of targets that can be engineered into personalized cancer vaccines and CAR-T-cells. OncoImmunity will work in collaboration with Ola Myklebost's group at the University of Bergen to validate the software on clinical samples from sarcoma patients. Once the NeoaAbScan software has been validated OncoImmunity will integrate the NeoAbScan platform with its current Immune-Profiler software which can accurately predict HLA-restricted neoantigens from NGS data. We envisage that the integration of the NeoAbScan platform into the Immune-Profiler software will enable OncoImmunity to penetrate the current target market more rapidly and enable the company to target new opportunities in the field of liquid biopsy-based diagnostics and the rapidly growing CAR-T-cell market - that are not directly targetable using the current Immune-Profiler platform.