Building multiscale computable model of Alzheimer's disease and identification of novel mechanisms for new therapeutic interventions

Abstract

Despite an avalanche of data in the field of biomedicine, we are obviously not managing well to extract meaningful information from this vast amount of data to better understand complex diseases and their mechanisms. Something must be wrong with the paradigmatic "let us generate new data", that drives current biomedical research. After all, we still have only a limited number of approved drugs available for many complex diseases; interpreting data and associating them with underlying molecular mechanisms of the disease is still a substantial challenge. Approaches that look into a wider perspective of the whole disease etiology as opposed to investigating on specific perturbed pathways or differentially expressed genes bear the potential to go beyond mere pattern identification. Biological networks help to achieve this goal acting as a platform to integrate heterogeneous data and a priori knowledge that may comprise various causal and correlational relationships among biological entities. These networks will lay the ground for the identification of disease mechanisms.<br /> This thesis presents a new formalism that integrate all combinations of interactions with various types of entities from different sources to understand how a single perturbance between two interactors can totally modify or amplify the changes of the whole system. As a use case, I have built the biggest computable mechanistic model of Alzheimer's disease (AD) in the course of this work. The first outcome is the identification of an early perturbed mechanism on AD based on interference with the neurotrophin signaling pathway. Secondly, I have linked SNP-associated effects to a larger functional context, which corroborates the comorbid association between AD and type 2 diabetes mellitus. Thirdly, I have systematically linked genetic and epigenetic alterations of DNA to the aetiology of diseases. Whilst the established computable model is specific to human pathophysiology, I have taken the opportunity of its existence to tackle one of the key questions of translational Alzheimer research, namely the functional equivalence of transgenic mouse models with the human disease pathophysiology. I compared the functional, mechanism inventory of a pre-clinical mouse model with the pathophysiology mechanisms that were described for humans in the area of neuro-inflammation. That analysis was extended towards pharmacology, where I analyzed – on the basis of the putative mechanism of action of a discontinued AD targeted drug; Celecoxib, - the reasons why that drug failed in the late phases of clinical trials. As I could show, the pre-clinical mouse experiments did not reflect the mechanistic context that is active in humans; which explains at mechanism-level the late failure of the drug despite promising results of the pre-clinical studies done with experimental animals. Lastly, I have used a comprehensive inventory of Alzheimer disease mechanisms to trace the investment of the pharmaceutical industry in AD drug development. I could demonstrate, how small the spectrum of candidate pathophysiology mechanisms is that the pharmaceutical industry is working on and I could show, how reluctant big pharma companies are to move from the "established targets" or "well-known pathways" into mechanisms that are novel, "ignored" or at least "not targeted" yet.