Computational modeling applications: protein - protein docking, hot spots prediciton and alloteric effects (aplicaciones de modelización computacional: interacción proteina - proteina, predicción de residuos responsables de esas interacciones y efectos alostéricos)

English summary: Protein-Protein Interactions (PPI) are key events in most of the essential processes that occur in living organisms. For example, signals coming from the exterior of a cell are sensed through PPI involving many signaling molecules. This cellular communication is very important in a large number of biological processes and can lead to disease when misregulated (as in cancer for example). Because of the importance of PPI, lots of efforts have been made, for many decades, to understand the rules governing these associations and several methods are currently available for their 3D structure determination such as X-ray crystallography, NMR and EM. However, only about 8800 structures of heterogeneous macromolecular complexes have been solved so far because of their difficulty. Consequently, computational approaches as docking are needed to model PPI in order to understand protein-protein association as a detailed energetic and structural knowledge of PPI, with the ultimate goal of designing drugs to modulate interactions of therapeutic interest. We participated to CAPRI (Critical Assessment of PRedicted Interactions), an international protein-protein docking experiment with our docking protocol called pyDock. CAPRI is a blind test to evaluate the ability of protein-protein docking algorithms to predict the binding-mode between two interacting proteins before the public release of their complexed structure and pyDock gave excellent success rates as compared to the most competitive docking methods participating in CAPRI. The pharmaceutical industry interest in PPI is huge because they are involved in virtually all essential biological processes and the potential applications to modulate specific PPI with small compounds spans over a large range. Thus, targeting PPI with small molecules is becoming the Holy Grail of drug discovery. The pharmaceutical industry has shown many examples of small molecules successfully designed to fit into enzyme active sites, but there are still very few cases of small compounds designed to target PPI. However, two strategies to modulate PPI are emerging: targeting protein-protein hot-spots and taking advantage of allosteric effects. Indeed, it is believed that targeting protein-protein hot-spots (residues directly responsible for the interaction from the energetic point of view) could help to the long-awaited goal of disrupting complexes with small molecules. We used our docking protocol called pyDock to predict hot-spots by calculating the Normalized Interface Propensities (NIP) of the 100 lowest energy poses on different datasets of complexes. We reached a quite good positive predictive value (78%) but a limited sensitivity. Interestingly our approach does not need any previous knowledge of the complex structure whereas most of the other does. A likely allosteric effect that could be used to develop new therapeutic strategies has been shown recently for the androgen receptor (AR), a transcriptional factor implicated in different human diseases as prostate cancer. We studied this nuclear receptor as well as 9 AR mutants (found in prostate cancer or androgen insensitivity syndromes) by molecular dynamics and described for the first time a clear allosteric effect between two different pockets: AF-2 and BF-3 that could explain the effects observed in vitro of several mutants. Resumen en castellano: La mayor parte de proteinas en la celula actuan en asociacion con otras proteinas, formando complejos especificos. Las interacciones proteina-proteina (IPP) estan implicadas en la mayoria de los procesos esenciales que ocurren en los seres vivos. Por ejemplo, todas las senales que provienen del exterior se transmiten al interior de la celula mediante interacciones proteina-proteina de moleculas senalizadoras. Esta comunicacion es muy importante en muchos procesos biologicos y puede desencadenar enfermedades cuando no esta regulada. Se dispone actualmente de una serie de metodos para la determinacion estructural de complejos, como la cristalografia de rayos X, la resonancia magnetica nuclear (RMN) y la microscopia electronica (EM). Sin embargo, solo hay alrededor de 8800 estructuras de complejos macromoleculares en el PDB dada la dificultad de resolverlos. En consecuencia, metodos computacionales de docking son necesarios para modelizar los IPP y entender la asociacion entre dos proteinas con el objetivo ultimo de disenar farmacos para modificar interacciones de interes terapeutico. Hemos participado en CAPRI (Critical Assessment of PRedicted Interactions), un experimento internacional de evaluacion de metodos de prediccion de IPP con nuestro metodo de Docking llamado pyDock. Este experimento permite la comparacion directa y objetiva del rendimiento de los diferentes metodos de docking y pyDock dio resultados altamente competitivos comparados con los de otros metodos. La industria farmaceutica esta muy interesada por las IPP dado que estan implicadas virtualmente en todos los procesos biologicos y las aplicaciones posibles para modificar IPP de interes terapeutico con pequenas moleculas son ilimitadas. La industria farmaceutica presenta numerosos ejemplos de moleculas de bajo peso molecular disenadas para unirse al sitio activo de enzimas, pero hay todavia muy pocos casos de compuestos capaces de modificar interacciones proteina-proteina, a pesar de que el interes en este campo es inmenso. Sin embargo, dos estrategias para modificar las IPP estan disponibles: usar los residuos hot-spots y aprovechar los efectos alostericos. Los residuos directamente responsables de la union entre proteinas (llamados hot-spots) podrian ayudar a modificar complejos mediante moleculas pequenas, lo que podria tener enormes aplicaciones biologicas y terapeuticas. Hemos explorado el uso de nuestros valores llamados NIP (por Propensidad Normalizada a la Interfase), derivados de la distribucion de orientaciones de docking mediante nuestro programa pyDock, para la prediccion de hot-spots. La principal ventaja de nuestro metodo es que, al contrario de otros previamente descritos, basta la estructura de las proteinas individuales. Permite predecir hot-spots con un valor predictivo positivo de 78% pero una sensitividad limitada. Recientemente se ha descubierto un efecto alosterico que podria ayudar al desarrollo de nuevas estrategias terapeuticas en el receptor de androgenos, un factor de transcripcion que tiene un papel significativo en una gran multitud de patologias humanas (por ejemplo en el cancer de prostata). Hemos estudiado este receptor nuclear junto con 9 de sus mutantes (encontrados en patologias humanas) por dinamica molecular y hemos descrito por primera vez un efecto alosterico claro entre dos bolsillos AF-2 y BF-3 que podria explicar los efectos observado in vitro por algunos mutantes.