Structural and dynamic modeling of molecular systems at different length scales

The continuous growth of computing power, both in terms of hardware and software resources, has made the computational (in-silico) approach to complex scientific problems a very profitable tool, which provides useful information to support, interpret or in some cases even reproduce the experimental datum from first principles. Methods have become cheaper and faster in the last two decades, thanks also to the development of more efficient algorithms, able to extract in full the computational power contained in novel hardware solutions (e.g. parallel computing and GPUs-based hardware), and to provide relatively easy-to-use software packages for diverse applications. Nowadays the computational approach is employed in several scientific areas, covering many different applied disciplines such as medicine, engineering, chemistry, physics, materials science and many others. In particular in this thesis work, some of the main approaches of computational chemistry, namely quantum mechanics, classical molecular dynamics and hybrid methods, are applied to the study of biomolecules and macromolecules, in order to investigate different aspects like structure, dynamics, energetics and in particular flexibility. In addition to the aforementioned methods we also explore a fluido-dynamic approach to describe and simulate microfluidic systems, focusing the attention on the reactivity of the systems studied. All these approaches are size-dependent and because they have different computational costs, their application should be limited to a reasonable size of the studied system. The profound difference in terms of cost/accuracy are discussed, providing a link between the different methodologies scales, in order to exemplify how information gathered at smaller length scale can be considered as an accurate starting point to perform simulations at larger spatial scales, in what is nowadays know popularly as multiscale modeling. The connection between the high accuracy/high cost and low accuracy/low cost methods is commented upon, to illustrate how a multiscale modeling approach can allow, in specific cases, to augment at the same time the accuracy of the data calculated and the size of the system simulated