Proteins and DNA are essential to life as we know it and understanding their function is understanding their structure and dynamics. The importance of the latter is being appreciated more in recent years and has led to the development of novel interdisciplinary techniques and approaches to studying protein function. Three techniques to study protein structure and dynamics have been used and combined in different ways in the context of this thesis and have led to a better understanding of the three systems described herein. X-ray crystallography is the oldest and still arguably most popular technique to study macromolecular structures. Nuclear magnetic resonance (NMR) spectroscopy is a not much younger technique that is a powerful tool not only to probe molecular structure but also dynamics. The last technique described herein are molecular dynamics (MD) simulations, which are only just growing out of their infancy. MD simulations are computer simulations of macromolecules based on structures solved by X-ray crystallography or NMR spectroscopy, that can give mechanistic insight into dynamic processes of macromolecules whose amplitudes can be estimated by the former two techniques. MD simulations of the model protein GB3 (B3 immunoglobulin-binding domain of streptococcal protein G) were conducted to identify origins of discrepancies between order parameters derived from different sets of MD simulations and NMR relaxation experiments.The results highlight the importance of time scales as well as sampling when comparing MD simulations to NMR experiments. Discrepancies are seen for unstructured regions like loops and termini and often correspond to nanosecond time scale transitions between conformational substates that are either over- or undersampled in simulation. Sampling biases can be somewhat remedied by running longer (microsecond time scale) simulations. However, some discrepancies persist over even very long trajectories. We show that these discrepancies can be due to the choice of the starting structure and more specifically even differences in protonation procedures. A test for convergence on the nanosecond time scale is shown to be able to correct for many of the observed discrepancies. Next, MD simulations were used to predict in vitro thermostability of members of the bacterial Ribonuclease HI (RNase H) family of endonucleases. Thermodynamic stability is a central requirement for protein function and a goal of protein engineering is improvement of stability, particularly for applications in biotechnology. The temperature dependence of the generalized order parameter, S, for four RNase H homologs, from psychrotrophic, mesophilic and thermophilic organisms, is highly correlated with experimentally determined melting temperatures and with calculated free energies of folding at the midpoint temperature of the simulations. This study provides an approach for in silico mutational screens to improve thermostability of biologically and industrially relevant enzymes. Lastly, we used a combination of X-ray crystallography, NMR spectroscopy and MD simulations to study specificity of the interaction between Drosophila Hox proteins and their DNA target sites. Hox proteins are transcription factors specifying segment identity during embryogenesis of bilaterian animals. The DNA binding homeodomains have been shown to confer specificity to the different Hox paralogs, while being very similar in sequence and structure. Our results underline earlier findings about the importance of the N-terminal arm and linker region of Hox homeodomains, the cofactor Exd, as well as DNA shape, for specificity. A comparison of predicted DNA shapes based on sequence alone with the shapes observed for different DNA target sequences in four crystal structures when in complex with the Drosophila Hox protein AbdB and the cofactor Exd, shows that a combined ”induced fit”/”conformational selection” mechanism is the most likely mechanism by which Hox homeodomains recognize DNA shape and achieve specificity. The minor groove widths for all sequences is close to identical for all ternary complexes found in the different crystal structures, whereas predicted shapes vary between the different DNA sequences. The sequences that have shown higher affinity to AbdB in vitro have a predicted DNA shape that matches the observed DNA shape in the ternary complexes more closely than the sequences that show low in vitro affinity to AbdB. This strongly suggests that the AbdB-Exd complex selects DNA sequences with a higher propensity to adopt the final shape in their unbound form, leading to higher affinity. An additional AbdB monomer binding site with a strongly preformed binding competent shape is observed for one of the oligomers in the reverse complement strand of one of the canonical (weak) Hox-Exd complex binding site. The shape preference seems strong enough for AbdB monomer binding to compete with AbdB-Exd dimer binding to that same oligomer, suggested by the presence of both binding modes in the same crystal. The monomer binding site is essentially able to compete with the dimer binding site, even though binding with the cofactor is not possible, because its shape is very close to the ideal shape. A comparison of different crystal structures solved herein and in the literature as well as a set of molecular dynamics simulations was performed and led to insights about the importance of residues in the Hox N-terminal arm for the preference of certain Hox paralogs to certain DNA shapes. Taken together all these insights contribute to our understanding of Hox specificity in particular as well as protein-DNA interactions in general.

In modern drug discovery, lead discovery is a term used to describe the overall process from hit discovery to lead optimisation, with the goal being to identify drug candidates. This can be greatly facilitated by the use of computer-aided (or in silico) techniques, which can reduce experimentation costs along the drug discovery pipeline. The range of relevant techniques include: molecular modelling to obtain structural information, molecular dynamics (which will be covered in Chapter 2), activity or property prediction by means of quantitative structure activity/property models (QSAR/QSPR), where machine learning techniques are introduced (to be covered in Chapter 1) and quantum chemistry, used to explain chemical structure, properties and reactivity. This thesis is divided into five parts. Chapter 1 starts with an outline of the early stages of drug discovery; introducing the use of virtual screening for hit and lead identification. Such approaches may roughly be divided into structure-based (docking, by far the most often referred to) and ligand-based, leading to a set of promising compounds for further evaluation. Then, the use of machine learning techniques, the issue of which will be frequently encountered, followed by a brief review of the "no free lunch" theorem, that describes how no learning algorithm can perform optimally on all problems. This implies that validation of predictive accuracy in multiple models is required for optimal model selection. As the dimensionality of the feature space increases, the issue referred to as "the curse of dimensionality" becomes a challenge. In closing, the last sections focus on supervised classification Random Forests. Computer-based analyses are an integral part of drug discovery. Chapter 2 begins with discussions of molecular docking; including strategies incorporating protein flexibility at global and local levels, then a specific focus on an automated docking program – AutoDock, which uses a Lamarckian genetic algorithm and empirical binding free energy function. In the second part of the chapter, a brief introduction of molecular dynamics will be given. Chapter 3 describes how we constructed a dataset of known binding sites with co-crystallised ligands, used to extract features characterising the structural and chemical properties of the binding pocket. A machine learning algorithm was adopted to create a three-way predictive model, capable of assigning each case to one of the classes (regular, orthosteric and allosteric) for in silico selection of allosteric sites, and by a feature selection algorithm (Gini) to rationalize the selection of important descriptors, most influential in classifying the binding pockets. In Chapter 4, we made use of structure-based virtual screening, and we focused on docking a fluorescent sensor to a non-canonical DNA quadruplex structure. The preferred binding poses, binding site, and the interactions are scored, followed by application of an ONIOM model to re-score the binding poses of some DNA-ligand complexes, focusing on only the best pose (with the lowest binding energy) from AutoDock. The use of a pre-generated conformational ensemble using MD to account for the receptors' flexibility followed by docking methods are termed “relaxed complex” schemes. Chapter 5 concerns the BLUF domain photocycle. We will be focused on conformational preference of some critical residues in the flavin binding site after a charge redistribution has been introduced. This work provides another activation model to address controversial features of the BLUF domain.