| The primary objective of the dissertation is the discovery of novel Acetylcholinesterase (AChE)
inhibitors and Positive Allosteric Modulators of Nicotinic Acetylcholine Receptors (nAChR)
using artificial intelligence (AI)-driven methodologies to address disorders related to synaptic
nicotinic system dysfunction. The study begins with a Structure-Activity Relationship (SAR)
study on Galantamine, a known AChE inhibitor, to identify structural requirements for inhibiting
AChE and positive allosteric modulation of α7 nAChR. Ten analogs, generated with the
systematic structural deconstruction of Galantamine, were synthesized and evaluated for their
activity on both receptors, highlighting cyclohexa(e)nyl ring and/or hydroxyl group as optimal
for activity. In the second part, we developed comprehensive, well-validated, interpretable, and
accessible machine-learning tools and platforms for predicting or identifying novel smallmolecule AChE inhibitors. The high performance of the best-derived models, which were
derived from different model types within each variant type, was observed and further validated
using cross-species datasets. Key structural components were identified through model
interpretations, and important AChE inhibitors were discovered by screening large compound
datasets. These findings were further validated with docking studies on selected compounds
intended for experimental validation. Finally, the models and the model development and
deployment process were made accessible through a user-friendly Streamlit web application. The
third part, AI-driven inverse design approaches, incorporating Tree-based Pipeline Optimization
Tool Automated Machine Learning (AutoML TPOT) and Local Interpretable Model-agnostic
Explanations (LIME), for the generation of new potential compounds with desired
xxiv
pharmacological properties, and chemical data augmentation, was presented. The emphasis of
the dissertation is that the synergy of AI tools and traditional drug discovery methods has the
potential to accelerate the identification of novel therapeutics, especially modulators of the
nicotinic acetylcholine system. Finally, challenges in AI-enhanced drug discovery of nAChR
PAMs, due to limited data, were presented, and the proposed solutions include novel approaches
developed in the realm of AI drug discovery, such as chemical data augmentation, few-shot
learning, and transfer learning, in combination with traditional computational approaches like
docking studies and pharmacophore modeling.
Keywords: artificial intelligence, nicotinic acetylcholine receptors, acetylcholinesterase, drug
discovery, inhibitors, positive allosteric modulation. |