Aggregate, automate, assemble

For over 60 years computers have been used to simulate biological systems in molecular detail using classical mechanics, using so-called molecular dynamics simulations. To perform a molecular dynamics simulation the system of interest has to be described using simulation input files, ensuring that the system is described using an appropriate model which reproduces the physical behaviour of the real system. Preparing these input files has, until now, only been possible for common, linear, polymers, such as proteins and DNA due to a historical focus on these systems. In chapter 2 we present software that is capable of setting up simulations for more complex systems such as branched or cyclic polymers using models at several resolutions. However, to assemble the required parameters for these polymers, building blocks are required. For atomistic molecular dynamics models, where every atom is represented as a separate particle, so-called automatic topology builders are commonplace. These tools can generate input parameters for arbitrary molecules automatically, without human intervention. For coarse-grained molecular dynamics models, where multiple atoms are grouped together to form a single, coarse, interaction site, these tools are not commonplace however. In chapter 3 we present the initial work to reformulate the problem of automatically creating coarse-grained input files as a clustering problem, which we aim to solve using machine learning techniques.Finally, in chapter 4, we present combined experimental and simulation work where we study an artificial peptide system, which is capable of self-replication, in an attempt to elucidate the self-replication mechanism