Fragment-Based AI for Antibiotic Discovery

Abstract

The threat of antimicrobial resistance is looming worldwide, highlighting the pressing need for innovative approaches to identify new antimicrobial agents. This paper reviews current strategies in which researchers are leveraging artificial intelligence (AI) techniques to accelerate the discovery of novel antibiotics and antibiotic classes. It highlights two key AI-driven strategies: (1) by repurposing of existing drugs using deep learning models like Chemprop, exemplified by the identification of the antibiotic Halicin, and (2) by de novo generation of new antibiotic candidates by computationally combining molecular fragments from known antibiotics, as can be performed by eSynth, which is a part of the AI-based DeepDrug pipeline. These complementary approaches showcase the ability of AI in efficiently navigating vast chemical spaces, uncovering structurally diverse antibiotics with distinct mechanisms of action, and ultimately revitalizing the antibiotic development process. By harnessing the power of AI alongside medicinal chemistry expertise, researchers are making important strides in addressing the global antibiotic resistance crisis.