Machine learning and tanimoto similarity based virtual screening of chemical libraries to develop novel PD therapeutics

Abstract

Parkinson's disease is a prevalent neurodegenerative disorder that affects a significant portion of the aging population worldwide. The etiology of Parkinson's disease involves the accumulation of misfolded proteins, particularly alpha-synuclein, leading to the formation of toxic aggregates in the brain. These aggregates are believed to contribute to the degeneration of dopamine-producing neurons in the substantia nigra, resulting in motor and cognitive impairments characteristic of the disease. Research in cheminformatics and computational chemistry offers a promising avenue for identifying small molecule compounds capable of modulating the aggregation of alpha-synuclein or mitigating its neurotoxic effects. By leveraging computational tools and databases of chemical compounds, researchers can screen and prioritize potential drug candidates with the desired properties for targeting Parkinson's disease pathology. Protein-ligand binding affinities play a crucial role in drug discovery and development. Accurately predicting these affinities can significantly streamline identifying potential drug candidates. In recent years, machine learning models have emerged as powerful tools for predicting protein-ligand binding affinities. The objective is to identify potential selective MAO-B and LRRK2 inhibitors among natural products and repurpose existing drugs for Parkinson’s disease treatment, employing LBVS and validating through docking and ADMET properties prediction. The use of high-quality benchmarking datasets, including a novel selectivity oriented dataset for MAO-B and LRRK2, is essential for both retrospective and prospective validation of virtual screening methodologies in this context. One such model is SVR, which has shown promising results in this domain. However, it is essential to thoroughly validate and benchmark to assess its reliability and performance compared to existing methods. Furthermore, the use of computational simulations enables the exploration of the interactions between small molecules and their target proteins at the molecular level, providing insights into the structural determinants of binding and potential mechanisms of action. We used DrugBank,PubChem, Real Enamine, and the Zinc Database to offer unique attributes conducive to the drug discovery process. This knowledge can inform the rational design and optimization of novel compounds with enhanced efficacy and safety profiles, offering new possibilities for therapeutic intervention in Parkinson's disease. Machine learning models trained on the training set achieved r squared value of MAO-B 69.0 and LRRK2 60.0. In the pursuit of discovering novel compounds targeting Parkinson's disease, the integration of cheminformatics and computational chemistry holds the potential to accelerate the identification and development of innovative treatments that address the underlying molecular mechanisms of the disorder.