Network Based Simulation on HPC for Translational Medicine: an Application to Anticoagulation

A machine learning investigation on Oral Anticoagulation Therapy (OAT) in elderly patients to improve the dynamic network modelization of the bioclinical process of coagulation cascade. We compared and tested techniques of supervised Machine Learning for classification and regression analysis, in order to develop an innovative model, compared to those currently available in the literature. The results of this investigation can provide adequate support for dual purpose: give an advantage in managing the therapy in patients with an increased risk of adverse drug events (ADRs), as elderly patients are; define pharmacodynamics characteristics of the drug employed in the oral anticoagulant therapy to finely configure a mechanistic model of involved pathway, providing further insights in therapy management. Therefore, simultaneously, we investigated and tested techniques of modeling and simulation in order to analyze the system in terms of both structure and dynamics to specify the qualitative and quantitative properties of the system. Such an approach would make it possible to profile a manageable model of the coagulation cascade which is able to adapt to the need for integration of recent pharmacological and genetic data, distinctive of modern pharmacogenomic characterization of anticoagulant therapy. Among the different mathematical modeling frameworks that have evaluated, the Temporal Extension of Petri Nets (Timed Petri Net), in particular the Stochastic Petri Net (SPN), seem to offer the most suitable characteristics. They permit an approach to modelization, combining an intuitive graphical representation to a valid solid mathematical, which give rise to a model numerically solving, able to represent the relational components and the quantitative time-dependent behaviors, including randomness, of a complex biological system such as coagulation cascade. The employment of such a framework, compared to the most popular models to differential equations provides a tool that can greatly enhance the interaction with biologists and clinicians enabling them to both evaluate the entire structure of system and cooperate for the development of any new modules. In order to simulate a high number of trajectories of the system, representative of several bioclinical scenarios, we employed HPC technology. The stochastic approach to simulate clinical events and interventions, due to its ability to represent the variability of the real system, could provide an interesting innovation for the integration of simulation techniques in clinical decision support according to recent translational medicine perspective.