Grant: Project Grant

Our current antibiotics have been discovered primarily by methods designed to identify molecules that inhibit the growth of a wide array of different bacterial species in standard lab conditions. However, despite many antibiotics that have been discovered - on the order of a hundred molecules used in the clinic - these can be grouped into just ~20 major structural classes and they all target a narrow range of cellular processes. This narrow range of antibiotic classes has contributed to antibiotic resistance. Furthermore, most current antibiotics with activity against bacterial pathogens (bad bacteria) also display antibacterial activity against commensal bacterial species that keep us healthy (good bacteria). Given the wide array of diseases that can result from disrupting commensal species, we must discover new antibiotics that spare them during treatment. We propose to apply cutting-edge artificial intelligence methods to design new species-specific antibiotics against one of the world's most burdensome bacterial pathogens, Acinetobacter baumannii. This research will lay the foundation for future efforts to design species-specific antibiotics against other difficult pathogens. Aim 1 - Artificial intelligence We will train neural network models to design antibiotics against A. baumannii. We plan to use types of models that can build molecules piece-by-piece until a chemical structure is achieved that has the properties of interest - in our case, activity against A. baumannii. Aim 2 - Lab validation After designing molecules with predicted activity against A. baumannii, a collection of these will be made in the lab and tested for A. baumannii-specific activity. The molecules that display this desired activity will then be studied to understand their function. Aim 3 - Infection models In cases where A. baumannii-specific molecules display strong activity in the lab, the best compounds will be tested in mouse models of infection to determine their efficacy.