Molecular Computational Drug Design Algorithms Through Machine Learning for Phytocompounds from Dimocarpus Longan

Since ancient times, plants have been used as a source of medicine. Today, increasingly commercially significant medications have their roots in plants. Medicinal plants have proven to treat various diseases with low or no side effects. This study analyzed the suitability of the drug design of the plant compounds from the Dimocarpus Longan (longan) plant for cancer treatment. Three-dimensional (3-D) of thirty selected bioactive compounds from the Longan plant were retrieved from the PubChem database. Retrieved plant compounds are screened using Lipinski’s Rule of Five, which provides a standardized requirement or criteria that a ligand should pass to be suitable for drug design. It establishes criteria for drug-like qualities and focuses on medication bioavailability. The Naïve Bayes Machine Learning algorithm was applied to the dataset to classify the plant compounds into two groups after screening using Lipinski’s Rule of Five. After classifying the plant compounds into two groups with optimal accuracy, the most influencing plant compounds were identified using principal component Analysis (PCA) techniques. Using Cross-Validation with k-fold=10 shows the accuracy produced by the Naïve Bayes Classifier, which reaches 93.33%. This study suggests that α-terpineol (PubChem ID: 442501) may be the safest and most effective cancer treatment. This in silico analysis of α-” ’terpineol’s anticancer properties will aid in creating a new and effective drug for cancer therapy.