Modelling of biomolecular complexes by data-driven docking

Abstract

Proteins play a major role in biology by interacting with each other and with other biomolecules. The study of these interactions is of fundamental importance to understand cellular processes, and this could be a key towards understanding mechanisms of diseases and possible development of drugs. As a result, considerable attention is given to the study of protein interactions. As described in the introduction of this thesis (chapter 1), a key task is to unravel these interactions at atomic details. However, traditional structure determination methods (X-ray crystallography and NMR) encounter serious experimental difficulties in dealing with protein complexes. Therefore, complementary computational methods are needed. This thesis describes the modelling of biomolecular complexes by data-driven docking, a computational method that, based on the known structures of the constituents of a complex and any information about their interface, derives a model for the structure of the complex. Chapter 2 reviews several approaches towards the use of experimental information in docking and introduces the data-driven docking method HADDOCK which we used in all the modelling described in this thesis. Chapters 3 and 4 focus on the use of NMR information in data-driven docking. Chemical shift perturbation (CSP) data allow assessment of the interface in a complex. These data can be combined with Residual Dipolar Couplings (RDCs, chapter 3) or diffusion anisotropy data (chapter 4), which provide complementary orientational restraints. In chapter 3 this methodology is applied towards the determination of the solution structure of the ubiquitin dimer. In Chapter 5 we turn towards the docking methodology itself. We describe a novel solvated docking protocol that accounts for water in the docking process. We demonstrate the feasibility of this approach and show that the docking results in general improve. Our protocol adds considerable information to existing docking approaches, since it allows to predict structural water molecules at an interface, which can be particularly valuable for drug design for example. The last two chapters are directed towards validation and application of our docking methodology. Chapter 6 describes the results of our participation to CAPRI (Critical Assessment of PRedicted Interactions). In CAPRI, participants must make blind predictions of the structure of a complex within a limited time; these are then compared by independent assessors to the yet unpublished experimental structure. Our participation to CAPRI was quite successful, and our overall performance places us in the top of the field. In the final chapter, HADDOCK is applied to the study of complexes along the cytochrome c oxidase copper-delivery pathway. The correct assembly of cytochrome c oxidase (which generates the proton gradient that drives synthesis of ATP, the major biological energy carrier) depends on this pathway. So far, structural details about interactions between the proteins involved in this pathway (cox17, sco1, cox2 and cox11) are unknown. Here, models for the structures of the human complexes cox17-sco1, sco1-cox2, cox17-cox11 and cox11-cox11 were generated. These models shed light on the path that copper travels during its way towards cytochrome c oxidase and provide new starting points for experimental studies.