dc.description.abstract
Proteins are the embodiment of the message encoded in the genes and they act as the building blocks and effector part of the cell. From gene regulation to cell signalling, as well as cell recognition and movement, protein-protein interactions (PPIs) drive many important cellular events by forming intricate interaction networks. The number of all non-redundant human binary interactions, forming the so-called interactome, ranges from 130,000 to 650,000 interactions as estimated by different studies. In some diseases, like cancer, these PPIs are altered by the presence of mutations in individual proteins, which can change the interaction networks of the cell resulting in a pathological state. In order to fully characterize the effect of a pathological mutation and have useful information for prediction purposes, it is important first to identify whether the mutation is located at a protein-binding interface, and second to understand the effect on the binding affinity of the affected interaction/s. To understand how these mutations can alter the PPIs, we need to look at the three-dimensional structure of the protein complexes at the atomic level. However, there are available structures for less than 10% of the estimated human interactome. Computational approaches such as protein-protein docking can help to extend the structural coverage of known PPIs.
In the protein-protein docking field, rigid-body docking is a widely used docking approach, since es fast, computationally cheap and is often capable of generating a pool of models within which a near-native structure can be found. These models need to be scored in order to select the acceptable ones from the set of poses. In the present thesis, we have characterized the synergy between combination of protein-protein docking methods and several scoring functions. Our findings provide guides for the use of the most efficient scoring function for each docking method, as well as instruct future scoring functions development efforts
Then we used docking calculations to predict interaction hotspots, i.e. residues that contribute the most to the binding energy, and interface patches by including neighbour residues to the predictions. We developed and validated a method, based in the Normalize Interface Propensity (NIP) score.
The work of this thesis have extended the original NIP method to predict the location of
disease-associated nsSNPs at protein-protein interfaces, when there is no available structure for the protein-protein complex. We have applied this approach to the pathological interaction networks of six diseases with low structural data on PPIs. This approach can almost double the number of nsSNPs that can be characterized and identify edgetic effects in many nsSNPs that were previously unknown. This methodology was also applied to predict the location of 14,551 nsSNPs in 4,254 proteins, for more than 12,000 interactions without 3D structure. We found that 34% of the disease-associated nsSNPs were located at a protein-protein interface. This opens future opportunities for the high-throughput characterization of pathological mutations at the atomic level resolution, and can help to design novel therapeutic strategies to re-stabilize the affected PPIs by disease-associated nsSNPs.
en_US
