Intestinal diseases caused by bacteria plague numerous people around the world, but the pathogens are increasingly resistant to antibiotics and difficult to overcome. Now, researchers from Tel Aviv University (TAU) have created an artificial intelligence platform that can identify the specific proteins that allow bacteria to infect the intestines – a technique that paves the way for the creation of smart drugs to neutralize the proteins and prevent disease without using antibiotics.

Doctoral student Naama Wagner and Prof. Tal Pupko, head of the School of Biomedicine and Cancer Research at TAU’s Faculty of Life Sciences and the new Center for Artificial Intelligence & Data Science conducted the research. The international partners in the study included researchers from Imperial College (led by Prof. Gad Frankel) and the Institute for Cancer Research, both in London, and the Technical University and the National Center for Biotechnology in Madrid. It has just been published in the prestigious journal Science under the title “Decrypting pathogen effector networks.” 

Such diseases are caused by pathogenic bacteria that attach to our intestinal cells. Once attached, the bacteria use a kind of molecular syringe to inject intestinal cells with proteins called “effectors” that work to control healthy cells – like hackers that take over computer servers using a combination of lines of code. But until now, scientists have not known what protein combination it is that cracks the cell’s defense mechanisms. The new artificial intelligence platform has identified novel effectors in the bacteria, which have been experimentally tested and validated. As a result, lab experiments conducted in London successfully predicted the protein combinations that lead to the pathogenic bacteria taking over the intestines.

“In this study, we focused on a bacterium that causes intestinal disease in mice, a relative of the E. coli bacteria that cause intestinal disease in humans, so as not to work directly with the human pathogen,” said Wagner. “The artificial intelligence we created knows how to predict effectors in a variety of pathogenic bacteria, including bacteria that attack plants of economic importance. Our calculations were made possible by advanced machine-learning tools that use the genomic information of a large number of bacteria. Our partners in England proved experimentally that the learning was extremely accurate and that the effectors we identified are indeed the weapons used by the bacteria.”

“Pathogenic bacteria are treated with antibiotics,” added Pupko. “But antibiotics kill a large number of species of bacteria in the hope that the pathogenic bacteria will also be destroyed. Antibiotics are thus not a rifle but a cannon. Moreover, the overuse of antibiotics leads to the development of antibiotic-resistant bacteria, a worldwide problem that is getting worse,” Pupko continued. 

“Understanding the molecular foundation of the disease is a necessary step in the development of drugs that are smarter than antibiotics, which will not harm the bacterial population in the intestines at all. This time we discovered the effectors of gut bacteria that attack rodents, but this is just the beginning. We are already working on detecting effectors in other bacteria in an attempt to better understand how they carry out their mission in the target cells they are attacking,” he concluded.