Molecular dynamics based methods for the computation of standard binding free energies and binding selectivity of inhibitors of proteins of pharmaceutical interest
The field of Computer Aided Drug Design (CADD) has experienced substantial developments over the last few decades thanks to a rapid growth incomputing power. In particular, Molecular Dynamics (MD) simulations and associated techniques have earned increased attention within the pharmaceutical sector thanks to their rising accuracy and diminishing cost. However, there are still limitations in the usage of these methods, due to thedifficulty of sampling the rugged energy landscapes of protein-ligand complexes. The main theme of this work is to address the sampling problem of MD methods for predicting the binding free energies of different biomolecular complexes. This work starts using MD simulations as a sampling technique for a relative free energy calculation protocol using the Sire Open Molecular Dynamics (SOMD) software. This protocol was then integrated in a ligand design workflow to optimize the binding selectivity of cyclophilin (Cyps) inhibitors. Cyps are proteins known to play a vital role in various diseases, such as cancer, Alzheimer and viral infections. Most Cyp inhibitors to date,however, are cyclic peptides that have potency in the nanomolar range but produce severe side effects, are complex to synthesize and display complex pharmacokinetic profiles. Thus, there is a need for new selective smallmolecules targeting specific Cyps isoforms, in order to gain new insightsfor the inhibition of these therapeutically vital proteins. The computational workflow was able to suggest auspicious designs that they will be synthesized and characterized using biophysical techniques from Alison Hulme’s lab. Following, MD simulation methods were employed for the more challenging task of predicting the absolute free energies of binding of protein-ligand complexes. For this purpose, an Alchemical Free Energy (AFE) protocol was generated and its efficiency was evaluated in the Statistical Assessment of Modelling of Proteins and Ligands (SAMPL6) challenge. SAMPL challenges involve a series of blinded predictions of standard binding freeenergies for toy host-guest molecules. The results obtained from our protocol were ranked among the top submissions in terms of accuracy and correlation with experimental data. Encouraged by these results, we wanted to compare the efficiency of the AFE protocol versus a Markov State Modelling (MSM) protocol for the calculation of the standard binding free energy of a ligand to the intrinsically disordered protein c-Myc. The oncoprotein c-Myc is overexpressed in over 70% of human cancers and its inhibition has been considered the holygrail in cancer therapy. Due to its structural elasticity it is difficult to perform structure-based drug design methods for the discovery of novel compounds. The results showed that MSM can describe accurately the binding process of the ligand to the oncoprotein c-Myc, but the binding free energies were similar with the ones of the AFE protocol. Finally, an adaptive sampling protocol was established for the computation of the standard binding free energy and binding selectivity of lead-like ligands for the flexible protein MDM2. MDM2 is a vital protein that acts as an inhibitory mechanism of the transcription factor p53. p53 plays animportant role in the regulation of cellular processes and suppression of tumor development. For this reason, it is important to develop methods for the discovery of novel ligands that could inhibit the MDM2-p53 interaction through binding to the MDM2 protein. The results of the adaptive sampling study were encouraging as the protocol was able to predict binding selectivity trends for the MDM2-ligand complexes approximately six times faster than the original AFE protocol.