Identifying new toxins using machine learning

New research at the Hebrew University reveals how bacteria use a special mechanism, like a tiny syringe, to inject toxins into other organisms. Using artificial intelligence, the researchers identified over 2,000 possible toxins that may be injected through the syringe, some of which could potentially be used as new antibiotics

Molecular Systems Biology portal showing eCIS injects AI-predicted toxins. Artist's impression by Dr. Yitzhak Yadgari
Molecular Systems Biology portal showing eCIS injects AI-predicted toxins. Artist's impression by Dr. Yitzhak Yadgari

New research reveals how bacteria use a special mechanism, like a tiny syringe, to inject toxins into other organisms. This is important because it helps us understand how bacteria communicate with their host and probably with other bacteria as well. Using artificial intelligence, the researchers identified over 2,000 possible toxins that may be injected through the syringe, some of which could potentially serve as new antibiotics. These findings, obtained through a combination of artificial intelligence and microbiology, can lead to new ways to treat infections and create new biotechnological tools.

Microscopic syringes and artificial intelligence: Scientists uncover new bacterial weapons

In research published on the cover of the August 2024 issue of the journal Molecular Systems Biology, researchers have revealed new secrets about a fascinating bacterial weapon system that works like a microscopic syringe. The team, led by Dr. Assaf Levy from the Hebrew University and in collaboration with researchers from the Hebrew University and the University of Illinois at Urbana-Champaign, made significant progress in understanding the extracellular injection system (eCIS), a unique mechanism that bacteria and archaea use to inject toxins into other organisms.

Cracking the bacterial code using artificial intelligence

The eCIS is a 100 nanometer long weapon that evolved from viruses that previously attacked microbes (phages). During evolution, these viruses lost their ability to infect microbes and became injectors that inject toxins into various organisms, such as insects. Previously, Levy's research group identified eCIS as a weapon carried by over 1,000 species of microbes. These microbes rarely attack humans, and the role of eCIS in nature remains largely unknown (Geller et al. 2021). However, we know it is loaded and injects protein toxins.

The specific proteins injected by the eCIS and their functions have long remained a mystery. Before the study, we knew of about 20 toxins that eCIS could load and inject. To solve this biological puzzle, the research team developed an innovative machine learning tool that combines genetic and biochemical data of various genes and proteins to accurately identify these elusive toxins. The project led to the identification of over 2,000 potential toxin proteins.

"Our discovery not only illuminates the way microbes communicate with their hosts and possibly with each other, but also demonstrates the power of machine learning in discovering new gene functions," explains Dr. Levy. "This could open new avenues for the development of antimicrobial treatments or biotechnological tools innovative."

New toxins with enzymatic activities against different molecules

Using AI technology, the researchers analyzed 950 microbial genomes and identified an impressive 2,194 potential toxins. Of these, four new toxins (named EAT14-17) were experimentally confirmed by proving their ability to inhibit the growth of bacterial or yeast cells. Impressively, one of these toxins, EAT14, was found to inhibit intracellular signaling in human cells, highlighting its potential to affect human health. The group showed that the new toxins probably act as enzymes that damage the target cells by damaging proteins, DNA or the critical molecule for energy metabolism. In addition, the group was able to decipher the protein sequence code that enables the loading of toxins into the eCIS syringe. Recently, it has been shown that eCIS can be used as a programmable injector that can be engineered for injection into various cell types, including brain cells (Krietz et al. 2023). The new findings from the current paper take advantage of this ability by providing thousands of toxins that are naturally injected by eCIS and the code that allows them to be loaded into the eCIS injector. The code can be transferred to other proteins of interest.

From microscopic approaches to medical breakthroughs

The research findings can have wide implications in the field of medicine, agriculture and biotechnology. The newly identified toxins may be used to develop new antibiotics or pesticides, efficient enzymes for various industries, or to engineer microbes that can target specific pathogens. This research highlights the enormous potential of combining biology with artificial intelligence to solve complex problems that can ultimately benefit human health.

"We are actually deciphering the weapons that bacteria have developed and continue to develop to compete for resources in nature," adds Dr. Levy. "Microbes are creative inventors and it is satisfying to be part of a group that discovers these amazing and surprising inventions."

The research was led by the students Alex Danov and Inval Polin from the Department of Plant Pathology and Microbiology, Institute of Environmental Sciences, in collaboration with Prof. Tommy Kaplan (School of Computer Science and Engineering) and Dr. Philippos A. Papathanos (Department of Entomology) from the Hebrew University of Jerusalem, and in collaboration with Prof. Brenda A. Wilson of the University of Illinois at Urbana-Champaign.

for the scientific article

More of the topic in Hayadan:

Leave a Reply

Email will not be published. Required fields are marked *

This site uses Akismet to filter spam comments. More details about how the information from your response will be processed.