A Classification of Antibiotic and Non-Antibiotic

Classification of Antibiotic and Non-Antibiotic Compounds in Moringa oleifera Using a Boosting Technique with the Decision Tree Algorithm

Abstract

Indonesia has high biodiversity, including medicinal plants such as Moringa oleifera. The process of identifying bioactive compounds as potential antibiotics generally requires considerable time and cost. Therefore, this study applies a machine learning approach to classify antibiotic and non-antibiotic compounds by using a Boosting technique on the Decision Tree algorithm, particularly under imbalanced class conditions. The dataset consists of 3,904 compounds as training data and 38 compounds from Moringa as prediction data. The results show that the Decision Tree model without Boosting achieved a recall of 0.6089 and a G-Mean of 0.7728, while the application of Boosting improved the recall to 0.7542 and the G-Mean to 0.8636. This indicates that the model becomes better at identifying the minority class. In addition, the prediction results on Moringa compounds identified three compounds as potential antibiotics, namely Niazicin A, Niazinin A, and [(2S,3R,4S,5S,6R)-3,4,5-trihydroxy-6-(hydroxymethyl)oxan-2-yl] (1E)-N-sulfooxy-2-[4-[(2S,3R,4R,5R,6S)-3,4,5-trihydroxy-6-methyloxan-2-yl]oxyphenyl]ethanimidothioate, with Niazicin A consistently classified by both models. These findings suggest that applying Boosting to Decision Tree can improve model performance and provide an initial insight for identifying antibiotic compounds from natural sources.