Abstract
Alzheimer’s disease (AD) is a permanent neurodegenerative disease listed as one of the leading causes of death in several countries. Common symptoms of AD include loss of memory, impaired speech and disorientation which may lead to death. A number hypotheses have been cited as the main reasons behind neurodegeneration in AD. However, deposition and aggregation of amyloid beta (Aβ) fibrils has been highlighted as the major culprit behind the pathogenesis of Alzheimer’s disease. To date, no cure has been reported for the disease and definitive diagnosis remains a huge setback in curbing this disease. Currently, only three drugs, flutemetamol, florbetaben and florbetapir, have been approved for clinical use. Even then, there still are major concerns about false or incorrect diagnosis. Several molecules have been proposed, however, they fail to make it through the drug discovery pipeline due to a variety of reasons that include, high toxicity, poor blood/brain barrier (BBB) penetration and low potency. As a result, this calls for a need to search for alternative, target specific molecules with minimal off-target interactions. The conventional way of drug design is very time consuming, expensive and generally resource demanding. However, computer aided drug design offers the best alternative to approaches of evaluating millions of compounds in a time-effective and less costly way. Therefore, the research presented in this thesis focuses on identifying new compounds intended for use in the diagnosis of Alzheimer’s disease. In this study, a computational approach using both ligand based and receptor based drug design is employed. Firstly, molecular docking and molecular dynamics simulations were performed using eight protein models namely 2BEG, 2MXU, 2MVX, 5OQV, 2NAO, 2M4J, 5KK3 and 2LMN so as to gather insight vii on crucial physicochemical properties required for binding of ligands to amyloid fibrils and to make a choice on what protein model to use for virtual screening studies. The selected crystal structures represent common fragments and full-length fibrils of different morphologies. The fragments selected are 2BEG (pentamer), 2MXU (dodecamer), 5KK3 (octadecamer) and 2LMN (dodecamer). For full-length 1-40 structures, one decamer (2MVX) and one nonamer (2M4J) were chosen. Finally, one hexamer (2NAO) and one nonamer (5OQV) were selected to represent full-length 1-42 amyloid beta fibril structures. 2MXU and 2MVX were isolated as the best candidates for virtual screening based on their stability during our investigative molecular dynamics simulations. Hydrophobic, π-π and hydrogen bond interactions were observed to be very important in the binding of ligands to fibrils. For the design/identification of novel molecules, a pharmacophore based virtual screening approach was employed and an atom based QSAR model was used to predict the activities of new compounds and lastly, all the molecules were computationally assessed for absorption, distribution, metabolism, excretion and toxicity (ADMET). Our best pharmacophore model, AHRR_1, successfully isolated 11 160 molecules from a database of over one million compounds. Further filtering with virtual screening yielded 12 molecules, 6 of which failed to pass the ADMET parameters required for a safe drug candidate. Therefore, the work presented in this thesis was useful in identifying new compounds which can be further tested experimentally or be used as starting materials for designing new compounds with improved affinity. In addition, the pharmacophore model and QSAR hypothesis generated can be extended to other databases in order to identify new compounds with activity towards amyloid beta fibrils.
Ph.D. (Chemistry)