Approved research

Deep neural modular genetic networks

Lay summary

Complex diseases are caused by the interaction of many factors, sometimes hundreds of them, as in one of our most interesting complex diseases, multiple sclerosis. These are both genetic, i.e. in the human DNA, and epigenetic, i.e. are influenced by the environment. Many of the factors and their interactions are not yet known. Because of this and other factors, most drugs for complex diseases are effective for less than half of the patients. A key limitation to understand complex diseases is that the underlying statistical assumptions most often are adaptations of the idea testing only a few factors at the same time for association to the disease. To make most of the emerging big data sets measuring several hundreds of thousands of factors simultaneously we will combine standard tools from artificial intelligence (AI) with our previously successful network medicine concepts that we previously published in well-reputed journals like Science Translational Medicine and Cell Reports. Artificial intelligence (AI) has recently shown enormous success in applications such as cancer diagnostics, self-driving cars, language translation, and board games. We want to explore whether those methods also could be used to improve our understanding of the human DNA. Our AI application will distil information from DNA and other biological data using deep genetic auto-encoders combined with biological network analysis. This information could increase the efficiency in other scientific studies of complex diseases. We will make the results available to the research community as computational processed drug discovery tools and present our findings to clinical collaborators and suggest potential new biomarkers. Thus, enabling increased precision in disease and drug association studies.