AGGLOMERATIVE HIERARCHICAL CLUSTERING ANALYSIS IN PREDICTING ANTIBACTERIAL ACTIVITY OF COMPOUND BASED ON CHEMICAL STRUCTURE SIMILARITY

Abstract

Resistance to antibiotics is increasing to alarmingly high levels. As antibiotics are less effective, more infections are becoming more complex and often impossible to treat. Numerous antibiotics discovered in marine organisms show that the marine environment, which accounts for over half of the world's biodiversity, is a huge source of novel antibiotics. This resource must be explored to identify next-generation antibiotics. This research aimed to predict antibacterial activity in marine compounds using a computational approach to reduce the cost and time of finding marine organisms, extracting, and testing numerous unknown marine compounds' bioactivities. We used a simple unsupervised learning approach to predict the biological activity of marine compounds, called agglomerative hierarchical clustering. We mixed antibiotic drug data in DrugBank Database and chemical compound data from marine organisms in the literature to compile our dataset. We applied five linkage methods in our dataset and compared the best method by assessing internal validation measurement. We found that the Ward with squared dissimilarity matrix is the best method in the dataset, and ten compounds from 73 compounds of marine compounds are determined as potential marine compounds which have antibacterial activity.