Dissecting Biomolecular Interactions by Integrative Modeling

Proteins and their intricate network of interactions are the mainstay of any cellular process. Dissecting their interaction networks at atomic detail is therefore invaluable, as this will pave the route to a mechanistic understanding of biological function. Atomic detail (high-resolution) information about structure and dynamics of biomolecular complexes is typically acquired by classical experimental methods such as X-Ray crystallography and NMR spectroscopy. Compared to other structural biology methods, these are the most accurate ones. They are, however, faced with many challenges, especially when the macromolecular systems under study become very large, comprise flexible or unstructured regions, exist in very tiny amounts, are membrane associated, or when their constituents interact only transiently. Due to the technical limitations of high-resolution methods and the other problems mentioned here, there is a considerable gap between the number of known 3D structures of macromolecular complexes and the amount of documented protein-protein interaction data. As a rescue strategy, structural biologists often resort to using different types of biochemical and biophysical experiments that can quickly provide accurate low-resolution information even for challenging systems. Most of the time, however, the collected data are rather sparse and/or of limited information content. These limitations call for integrative computational tools, like for example docking, that can, using some kind of physical model, judiciously combine and accurately translate sparse experimental data into structural information. In this thesis, we present the current state of integrative approaches, the challenges of the field and the recent development made regarding our information-driven docking program HADDOCK, in order to push back the limits of its applicability in integrative modeling. To that end, in Chapter 2, a method to model generic multi-body complexes by simultaneous docking of all components is described and its performance is depicted for six multimer complexes, composed of five symmetric protein homo-oligomers and one symmetric protein–DNA complex. In Chapter 3, we discuss an efficient divide-and-conquer approach, built on the multi-body docking ability of HADDOCK described in Chapter 2, to deal with large conformational changes upon binding. The performance of this approach is benchmarked on a set of 11 dimeric protein–protein complexes, covering a vast range of conformational change from 1.5 Å to as much as 19.5 Å. In Chapter 4, the usefulness of low-resolution shape data for scoring docking decoys is explored. A new scoring function is introduced that combines the regular HADDOCK score with the low-resolution shape information from either Small Angle X-ray Scattering or Collision Cross Section data. In Chapter 5, the methods developed in this thesis are demonstrated on two challenging real-case examples, for which different types of experimental data area available. Finally, the thesis ends with a Perspectives section on the future and limitations of integrative modelin