Incorporating prior knowledge to exploit public omics data in finding higher order genetic interactions

Abstract

Cancer is one of the most common causes of death in high-income countries with an estimated 4 million new cases and 1,9 million cancer-related deaths annually in Europe alone. Currently, the most common treatment approach is chemotherapy. This approach does not only damage tumour tissue, but healthy tissue as well, leading to numerous side effects. Targeted treatment options, which only target tumour tissue, have already proven to be a good alternative to chemotherapy. These treatments are, however, usually cancer specific and are only applicable if certain mutations are present. It is therefore of the utmost importance to continue researching new potential therapeutic targets. Genetic interactions could be used to uncover new targets. Genetic interactions are a phenomenon in which the mutation of two genes leads an unexpected phenotype than would be expected when comparing it to a singular mutation of one of the two genes. It has however been found that the complexity of genetic interactions is too great to be explained only by the interaction of two genes. It is thus beneficial to create a workflow which allows for the detection of higher order genetic interactions in cancer. In this study we propose a new approach to applying epiNEM on publicly available cancer data. The workflow is composed of three sections: the creating of patient subsets, differential expression analysis through a leave-one-out approach, and the significance estimation of epiNEM’s results. With this new approach we have been able to identify two significant interactions: the masking of NPM1 by DNMT3A which is modulated by FLT3 and the masking of FLT3 by NPM1 which is modulated by DNMT3A. These interactions may aid in better understanding the complexities of acute myeloid leukaemia development and progression. In a follow up study, we need to test the validity of our workflow in vitro through a perturbation study of our MOIs.