Integrated virtual screening, molecular modeling and machine learning approaches revealed potential natural inhibitors for epilepsy.

Epilepsy, a prevalent chronic disorder of the central nervous system, is typified by recurrent seizures. Present treatments predominantly offer symptomatic relief by managing seizures, yet fall short of influencing epileptogenesis. This study endeavored to identify novel phytochemicals with potential therapeutic efficacy against S100B, an influential protein in epileptogenesis, through an innovative application of machine learning-enabled virtual screening. Our study incorporated the use of multiple machine learning algorithms, including Support Vector Machine (SVM), k-Nearest Neighbors (kNN), Naive Bayes (NB), and Random Forest (RF). These algorithms were employed not only for virtual screening but also for essential feature extraction and selection, enhancing our ability to distinguish between active and inactive compounds. Among the tested machine learning algorithms, the RF model outshone the rest, delivering an impressive 93.43 % accuracy on both training and test datasets. This robust RF model was leveraged to sift through the library of 9,000 phytochemicals, culminating in the identification of 180 potential inhibitors of S100B. These 180 active compounds were than docked with the active site of S100B proteins. The results of our study highlighted that the 6-(3,12-dihydroxy-4,10,13-trimethyl-7,11-dioxo-2,3,4,5,6,12,14,15,16,17-decahydro-1H cyclopenta[a] phenanthren -17-yl)-2-methyl-3-methylideneheptanoic acid, rhinacanthin K, thiobinupharidine, scopadulcic acid, and maslinic acid form significant interactions within the binding pocket of S100B, resulting in stable complexes. This underscores their potential role as S100B antagonists, thereby presenting novel therapeutic possibilities for epilepsy management. To sum up, this study's deployment of machine learning in conjunction with virtual screening not only has the potential to unearth new epilepsy therapeutics but also underscores the transformative potential of these advanced computational techniques in streamlining and enhancing drug discovery processes.