Protein-protein interaction networks for precision oncology

The limited understanding of phenotype-generating mechanisms at the molecular level is a key problem of medical research. Although large amounts of genetic data have been generated, it remains difficult to relate genomic lesions to disease phenotypes and their spatiotemporal aspects. Missense mutations derived from common genetic variation or acquired somatically can influence the function of protein complexes in specific tissues. We propose that combining the molecular mechanistic perspective on the data with advances in network biology offers new opportunities to study the causes of such complex genetic modification-based effects. The key objective of the project is to enable personalized cancer prognosis through a systems medicine approach. Based on the well-established principle that protein structure and, in particular, topology and structure of protein complexes determine the functional capabilities of cells and tissues, we will address a timely and urgent need to link genetic data to molecular and clinical phenotypes that depend on the function of protein complexes in a tissue-specific manner. The project aims at the identification of tissue-specific and tumor-specific computational models, in which patient-specific genetic variation influences protein interactions in ways that affect health. Such models will be initially derived by bioinformatics predictions, and then iteratively refined based on validation data generated by proteomics and genetics experiments. Specifically, predictive techniques developed in the project will be subjected to rigorous experimental verification by generating the corresponding protein complexes and using cross-linking experiments to determine whether and how they differ in topology and structure from the wild type. The models will be validated using samples from a well-annotated cohort of triple-negative breast cancer (TNBC) patients. The project proposes to generate several important data resources, analysis methods, software tools and services: - Methods for predicting disease mutations affecting protein-protein interactions (PPIs) between both globular and transmembrane proteins, leading to loss or gain of function. - Curated datasets of tissue- and tumor-specific interaction networks impacted by sequence variants. - Methods for creating interaction networks that take into account isoforms as well as dynamic and concentration-dependent aspects of PPIs. - Methods and software tools to detect structural and compositional alterations in protein complexes induced by genomic lesions - Curated protein complexes and cancer specific networks relevant for tumor-immune cell interactions in solid cancers - Experimental validation by chemical cross-linking and mass spectrometry of predicted structural or compositional alterations induced by genomic variation in a set of protein complexes. - Primary data and results from the immunogenomic analyses from the TNBC cohort used for validation. Compared to many correlation-focused biostatistical method developments, the proposed research program is unique in that it intends to identify and leverage mechanistic causal associations for highly reliable and generally applicable predictions.

In summary, the results of the project contributed considerably to the advancement of the field. Specifically, the development of a novel computational tool quanTIseq represents an important contribution to the computational toolbox for dissecting tumor-immune cell interactions from RNA-seq data. Over and above, we were able to develop a novel computational experimental approach for reconstructing signaling networks using tumour-derived organoids, perturbation experiments, and quantitative phosphoproteomics to construct signalling networks. Our network biology approach takes advantage of recent developments of the organoid model system, improvements in mass spectrometry-based proteomic technologies, and available knowledge resources.