Machine learning approaches to predict Acetylcholinesterase inhibition

Summary Acetylcholine (Ach) is a neurotransmitter at neuromuscular junctions and synapses in the autonomic and central nervous systems. It also functions as a signaling molecule in non-neuronal contexts related to cellular functions, such as proliferation and differentiation, as well as performing organ functions, like wound healing in skin or mucus production in lungs. Organophosphorus (OP) are one of the most common causes of poisoning worldwide. There are nearly 3 million poisonings per year resulting in three hundred thousand deaths of these approximately 8000 are in the USA. Because of their unique chemical properties, OPs bind to acetylcholinesterase (AChE), rendering the enzyme incapable of hydrolyzing ACh in the cholinergic synapses and neuromuscular junctions. Subsequent accumulation of ACh leads to overstimulation of the affected neurons acting through muscarinic and nicotinic receptors. The peripheral effects of excess systemic ACh include observable toxic signs (e.g., miosis, lacrimation, salivation, fasciculation, tremors and convulsions), as well as life- threatening cardiovascular and respiratory distress. Simultaneous progression of the cholinergic crisis within the central nervous system ultimately induces a state of unremitting seizure known as status epilepticus. Unmitigated OP-induced SE is associated with wide- spread neuronal damage, and concomitant cognitive and behavioral deficits. Besides the effects directly in humans, OPs can reach humans indirectly via expose to various types of organisms that have themselves been contaminated in the environment. Some of the adverse effects of pesticides on non-target organisms such as fish, amphibians and humans have also occurred as a result of biomagnifications of the toxic compounds. What is missing across public “Structure Activity/toxicity Relationship” databases are accessible machine learning models for scientists to use to extract knowledge from the small molecule data that is accumulating. We would propose predicting AChE inhibition from structure of the molecule alone. Our mission is therefore to make the various public datasets much more readily accessible to machine learning modeling by providing the underlying datasets ready to model as well as apply prebuilt models of our own. This project therefore covers automated curation, data integration and will build a research pipeline for machine learning model development for AChE inhibition. We now propose auto-curation of public AChE databases which use predominantly small molecule / biological activity data (such as IC50, Ki, EC50, or % inhibition etc), sorted by target and species. We will develop software to autocurate data, build machine learning models and identify potential molecules that inhibit AChE from human and other species in order to predict poisoning and possible environmental contamination. We will also validate these models with literature data outside of the training sets and understand the applicability domain of these models to other classes of molecules besides OPs. Our ultimate goal will be to provide software and models to predict AChE inhibition which will be a commercial product.