Abstract:
Despite the considerable progress in understanding cancer biology and cancer development that has been made over the last decades, the treatment options for cancer are still insufficient. This can be attributed to the tremendous heterogeneity of cancers, with respect to appearance, clinical outcome, and underlying genetic alterations. In traditional concepts of drug design and drug administration, pathologically similar diseases are treated with the same drugs. These approaches are not adequate to face the complexity of cancer. Personalized or individualized approaches, targeting individual characteristics of tumors, are promising concepts to develop successful treatment options for cancer with little side-effects.
The human organism is equipped with a powerful system that is capable of targeting abnormal cells specifically and efficiently: the immune system. T cells can distinguish healthy cells from infected or aberrant cells by scanning peptides that are presented on the surface of other cells. Genetic alterations in cancer cells can lead to the presentation of cancer-specific peptides that drive a very specific immune reaction against the cancer cells. These peptides are called cancer-specific T-cell epitopes. Each patient’s immune system is individual with respect to the peptides that can elicit an immune response. The design of tailor-made immunotherapies against individual tumors can thus be realized by using sets of patient- and tumor-specific T-cell epitopes in so-called epitope-based vaccines.
A first major challenge in the development of such individualized therapies lies in the analysis of genetic information of individual cancers, which is necessary to detect cancer-specific mutations. A second challenge is the correct identification and selection of T-cell epitopes resulting from these mutations. In this thesis, we present computational methods that address these challenges. Starting from next-generation sequencing data of cancer and normal tissue from individual patients, we identify those mutations that are uniquely present in the tumor. We integrate information from gene expression, biological pathways, and functional annotation of genes and proteins to select suitable mutations. These mutations form the basis for potential targets for individualized immunotherapies. We present prediction algorithms based on machine learning approaches that identify T-cell epitopes that are specific for a patient’s tumor and immune system. In order to bring the computational methods to clinical applications, results have to be obtained in a reliable, reproducible, and timely manner, and have to be made available to clinical researchers in an easy-to-use and intuitive way. An additional focus of this thesis is thus the development of pipelines, tools, and user-interfaces that facilitate a close integration between the computational analysis with the experimental application in a clinical setting.
We apply the presented methods to clinical data. The results show that a combination of high-throughput data, computational data analysis, and accurate prediction methods with clinical research can promote the development of new individualized treatment options for cancer.