In Silico Screening of Natural Compounds for Neuroprotection

Abstract

Neurodegenerative disorders -- Alzheimer's disease (AD), Parkinson's disease (PD), amyotrophic lateral sclerosis (ALS), and Huntington's disease (HD) -- collectively affect more than 50 million people worldwide, with prevalence rising steeply as global populations age. Despite decades of intensive pharmaceutical research, disease-modifying therapeutics remain unavailable for most of these conditions, and the clinical failure rate of small-molecule neuroprotective drug candidates exceeds 99%. Natural compounds -- the secondary metabolites of plants, fungi, and marine organisms - represent an underexplored pharmacological reservoir with established biological activities across anti-inflammatory, antioxidant, mitochondrial-protective, and protein aggregation-inhibitory pathways that are directly implicated in neurodegeneration. In silico screening -- applying computational molecular docking, pharmacophore modelling, ADMET prediction, and machine learning activity scoring to large natural compound libraries -- enables rapid prioritisation of candidates with the best predicted neuroprotective profiles for experimental validation. This study screens 12,847 natural compounds from COCONUT, SuperNatural III, and ZINC Natural Products databases against seven validated neuroprotective targets -- BACE1, AChE, MAO-B, alpha-synuclein aggregation, SOD1 misfolding, HDAC6, and NLRP3 inflammasome -- using a five-stage in silico pipeline. Top-ranked candidates include curcumin analogues (BACE1; AUC 0.87), epigallocatechin gallate (alpha-synuclein aggregation inhibition; IC50 pred = 0.84 µM), and withaferin A (HDAC6; docking score -9.4 kcal/mol). A Natural Compound Neuroprotection Score (NCNS) integrating target binding, CNS penetration, metabolic stability, and toxicity predictions is developed and applied to all 12,847 compounds, yielding 247 priority candidates for experimental validation.