Grant: Project Grant

Antibiotic resistance is a serious threat to public health worldwide. The situation is that antibiotic-resistant bacterial infections are increasing while discoveries of new antibiotics are decreasing. One of the reasons for the lack of new antibiotics is that with existing methods of antibiotic discovery, the chance to find a new drug that kills bacteria is extremely low, which discourages pharmaceutical companies to pursue antibiotic discovery. Here we will show that we can increase the chance of finding new antibiotics if we use artificial intelligence to predict antibiotic activity among the millions of drugs available. While most of the known drugs do not have antibiotic activity, some of them can kill bacteria with mutations in antibiotic targets, revealing a potential antibiotic activity that can be chemically improved. We will then analyze the growth of a large collection of bacterial mutants in antibiotic targets against a library of chemical molecules, including antibiotics. This experiment will show characteristic mutant responses for each molecule (fingerprints) where different mutants may be killed, indicating potential antibiotic activity. Then we will describe the shape and charge of the chemical molecules (structure) using a language that computers can understand. With this information, we will train an artificial intelligence system to identify which are the molecule structures that produce fingerprints characteristic of potential antibiotic activity. Once the artificial intelligence tool is built, we will use it to predict the potential antibiotic activity of millions of "virtual" drugs, based on their structure. We will confirm the artificial intelligence predictions by analyzing the fingerprints and the antibiotic activity of the selected drugs in our laboratory. Our research will facilitate the discovery of new antibiotics alleviating the current crisis of antibiotic-resistant bacterial infections.