An Approach to Identify Mechanism of Actions of Predicted Natural Antibiotics

Authors:

Abstract
Antibiotic resistance is a mounting global health challenge, demanding accelerated strategies for discovering new drugs and understanding their mechanism of action (MoA). In this study, we propose a computational framework to predict the MoA of natural antibiotics by leveraging Tanimoto molecular similarity metrics and network-based clustering. A curated dataset of known antibiotics with established MoA's was used to generate a similarity network, into which predicted natural antibiotics were integrated. Using the DPClus algorithm, we identified structurally coherent clusters that frequently aligned with known functional categories. The results suggest that predicted antibiotics sharing structural clusters with characterized compounds are likely to exhibit similar biological activities. Our approach provides an interpretable and scalable method for prioritizing novel antibiotic candidates for further validation, contributing to faster and more informed drug discovery pipelines.