Abstract
It is well established in the field of immunology and oncology that tumours express antigens unique to the cancer, called neoantigens, that can be recognized by T-cells which target the cancer cell for destruction. The expression of antigens, depends on several regular cell mechanisms like gene expression, the antigen processing pathway, and MHC binding. Cancer’s unique mutation profile contributes to the formation of mutated proteins which have the potential of producing neoantigens. By exploiting the neoantigen expression of cancer, immunotherapies can be developed, harnessing the biological processes of the adaptive immune system to attack the tumour. Each tumour’s unique and unpredictable mutation profile poses a big challenge for the treatment of sequencing data coming from Next Generation Sequencing methods. With the set of somatic mutations, one can ask which is the set of potential neoantigens that the tumour expresses and with which frequency? The answer to this question has the potential of contributing significantly to the creation of immunotherapies, targeting both unique and shared neoantigens, shared among a population of cancer patients. The creation of shared neoantigen therapies have the extra benefit of being more cost effect and readily available than patient specific ones. To answer the question, this project had the following goals: 1) integrate tools for DNA variant annotation that can detect peptide changes; 2) integrate MHC binding prediction tools; 3) use these tools with two gold standard databases of somatic variants, COSMIC and My Cancer Genome (MCG), and the list of cancer driver genes from the Cancer Gene Census (CGC), to study the frequency of neoantigens and detect patterns of neoantigens among known somatic cancer variants and known cancer driver genes. The end result was a framework to identify and prioritize shared neoantigens in cancer, by implementing a variant annotation and MHC binding prediction pipeline, and a probability model to select common neoantigens in populations. Applied to COSMIC and CGC, this framework revealed interesting trends, like the high load of potential MHC binder peptides of cancer driver genes such as TP53 gene, and the relative high MHC binding potential of the SLC34A2 gene. Applied with mutation frequency data from the MCG, this framework produced a visualization of how each MCG variant and gene affected the probability of shared neoantigens with a given HLA context, and a visualization of potential benefit of shared immunotherapies targeting shared neoantigens.