A Bacterial Immunity Machine Learning System
Peter C. DeWeirdt, Emily M. Mahoney, Michael T. Laub
Bacteria, like humans, can get sick. Bacteriophages, viruses that specifically target bacteria, infect a variety of bacterial species. After many years of fighting viral infections, the bacteria evolved ways to fight back by making their own defense systems. Over the years, humans have leveraged bacterial immune systems for their own biotechnology, giving rise to powerful technologies such as CRISPR-Cas9, a pair of molecular scissors that can cut DNA and modify it. These innovations in genetic engineering allow us to make many life-saving products from bacteria, like insulin for diabetes.
However, we don’t know all of bacteria’s tricks yet – which is why scientists have developed DefensePredictor, a powerful machine learning model that identifies proteins that may be related to a bacterium’s defensive systems. When this powerful system was applied to a variety of bacterial genomes, they predicted more than 5000 new prospective defense proteins. This gives us insight into components of the bacterial immune system and potential targets in medicine. In addition, this model was released as ‘open source’, meaning that other researchers can use the same base model to make even further discoveries. Ultimately, tools like this not only deepen our understanding of bacterial immunity but also pave the way for new antimicrobial strategies in an era of rising antibiotic resistance.
