In Silico Approaches for Polypharmacological Drug Design

Rational drug design experiences a shift from single-target drugs towards drugs that can deliberately interact with multiple biological targets in the human body. This approach is deemed more suitable for the treatment of complex diseases such as cancer or central nervous system disorders. A complex disease is most likely based on more than one biological target which interact with each other through a network. Highly selective drugs against a single biomolecule can be rendered ineffective by redundant pathways and fail-safes emerged from evolution. The process of drug design already utilizes computerized or in silico methods from the fields of systems biology and cheminformatics. These methods have to be tailored to the finding of molecular compounds with multi-target properties to adapt to the aforementioned change of paradigm. This thesis presents several in silico approaches applicable in polypharmacological drug design for the identification of one or more disease-related targets, the search for lead structures, and the optimization of the chosen lead structures. These methods comprise the simulation and parameterization of biochemical models by means of ordinary differential equation (ODE) systems, virtual screening based on structured support vector machines (SVMs), and quantitative structure-activity relationship (QSAR) models created with multi-task learning and support vector regression (SVR).