Modeling protein dynamics and protein-drug interactions with Monte Carlo based techniques

Abstract

[eng] A complete understanding of complex formation between proteins and ligands, a crucial matter for pharmacology and, more in general, in biomedicine, requires a detailed knowledge of their static and dynamic atomic interactions. The main objective of this thesis is to test recent developments in conformational sampling techniques in providing such a dynamical view. We aim at developing new protocols and methods for such a study. Moreover, we want to show how its application can aid in addressing existing problems in the biophysics of protein ligand interactions. Moreover, we apply and refine novel computational approaches aiming at a comprehensive description of the protein and protein-ligand energy landscape, progressing into the rational design of new inhibitors for particular targets. We provide here a summary of the main results. PELE was used for induce fit docking in protein kinases, mammalian target of rapamycin (mTOR) and BCL-2 family protein, particularly MCL-1 protein. Results produced a detailed atomic description of the binding modes of ligand/drug to the selected target. Overall, these results provide new data to understand the mechanism of action of these molecules, and provide new structural data that will allow the development of more Specific inhibitors for cancer treatments. Importantly, we demonstrate the critical role of sampling the protein-ligand dynamics in order to improve the docking score. Moreover, the findings reported here clearly shown the capabilities of PG (and its derivatives) for use in particular apoptotic targets. Following the previous goal, we aim at the implementation of the atomic detailed knowledge into the rational design of new inhibitors, aiming to enhance specificity and binding strength. Motivated by our success with validation studies (applied to several systems for protein-ligand interaction and induce fit procedure) we attempted to design a new inhibitor for a specific target. For doing so, we used the system from our second study: Molecular interactions of prodiginines with the BH3 domain of BCL-2 family members. We have shown how PELE can be used in effectively design improved compounds with significant better docking results. The PELE was applied to steroid Nuclear Receptors to unbiased simulations, where substrate/ligands were placed in the active site, to freely move through the protein and finding the channels, or outside the receptors allowed ligand to freely explore the protein surface. In this study, we demonstrated the applicability of the PELE method in solving relevant biophysical problems. In particular, using PELE we introduced a new structural and dynamic paradigm for ligand binding in steroid nuclear receptors. Using PELE, we create a protocol involving sequence comparison and all­atom protein-ligand induced fit simulations to predict PR resistance at the molecular level. We introduced a significant advance in predicting the affinity of different drugs against HIV-1 protease with several mutations. This study shows how computational techniques are capable of quantitatively discriminating resistance variants of HIV-1 protease. This application is fully automated and installed on PELE web server. Beside these main objectives based on methods application, we aim to add methodological improvements derived from the application and validation studies. We performed method development and studied PELE protocols to model long-time protein dynamics by means of normal mode perturbation and constrained minimization. New backbone perturbation combined with normal modes increased the capability of PELE method to explore local dynamics and large conformational changes.