Mapping biophysics through enhanced Monte Carlo techniques

• Autores: Israel Cabeza de Vaca López
• Directores de la Tesis: Manel Canales Gabriel (dir. tes.), Victor Guallar Tasies (dir. tes.)
• Lectura: En la Universitat Politècnica de Catalunya (UPC) ( España ) en 2015
• Idioma: inglés
• Tribunal Calificador de la Tesis: Juan Jesús Pérez González (presid.), Modesto Orozco López (secret.), Martin Zacharias (voc.)
• Programa de doctorado: Programa Oficial de Doctorado en Física Computacional y Aplicada
• Materias:
  • Física
    • Física molecular
  • Ciencias de la vida
    • Biofísica
• Enlaces
  • Tesis en acceso abierto en: TDX
• Resumen

  • This thesis is focused on the study of molecular interactions at the atomistic detail and is divided into one introductory chapter and four chapters referencing different problems and methodological approaches. All of them are focused on the development and improvement of computational Monte Carlo algorithms to study, in an efficient manner, the behavior of these systems at a classical molecular mechanics level. The four biophysical problems studied in this thesis are: induced fit docking between protein-ligand and between DNA-ligand to understand the binding mechanism, protein stretching response, and generation/ scoring of protein-protein docking poses. The thesis is organized as follows: First chapter corresponds to the state of the art in computational methods to study biophysical interactions, which is the starting point of this thesis. Our in-house PELE algorithm and the main standard methods such as molecular dynamics will be explained in detail. Chapter two is focused on the main PELE modifications to add new features, such as the addition of a new force field, implicit solvent and an anisotropic network specific for DNA simulation studies. We study, compare and validate the conformations generated by six representative DNA fragments with the new PELE features using molecular dynamics as a reference. Chapter three is devoted to applying the new methods implemented and tested in PELE to study protein-ligand interactions and DNA-ligand interactions using four systems. First, we study the porphyrin binding to Gun4 protein combining PELE and molecular dynamics simulations. Besides, we provide a docking pose that has been corroborated by a new crystal structure published during the revision process of the submitted study showing the accuracy of our predictions. In the second project, we use our improved version of PELE to generate the first structural model of an alpha glucose 1,6-bisphosphate substrate bound to the human Phosphomannomutase 2 demonstrating that this ligand can adopt two low-energy orientations. The third project is the study of DNA-ligand interactions for three cisplatin drugs where we evaluate the binding free energy using Markov state models. We show excellent results respect another free energy methods studied with molecular dynamics. The last project is the study of the daunomycin DNA intercalator where we simulate and study the binding process with PELE. Chapter four is focused on the computational study of force extension profiles during the protein unfolding. We added a dynamic harmonic constraint following a similar procedure applied in steered molecular dynamics to our Monte Carlo approach to fix or pull some selected atoms forcing the protein unfolding in a defined direction. We implement and compare with steered molecular dynamics this technique with Ubiquitin and Azurin proteins. Moreover, we add this feature to a well-known algorithm called MCPRO from William Jorgensen¿s group at YALE University to evaluate the free energy associated to the unfolding of the deca-alanine system. Chapter five corresponds to the introduction of a multiscale approach to study protein-protein docking. A coarse-grained model will be combined with a Monte Carlo exploration reducing the degrees of freedom to generate thousands of protein-protein poses in a quick way. Poses produced by this procedure will be refined and ranked through a protonation, hydrogen bond optimization, and minimization protocol at the all-atom representation to identify the best poses. I present two test cases where this procedure has been applied showing a good accuracy in the predictions: tryptogalinin and ferredoxin/flavodoxin systems.