McMaster University researchers have developed a new generative artificial intelligence (AI) model that can significantly speed up drug discovery, and have already designed an entirely new antibiotic in early tests.
This discovery demonstrates how AI can dramatically improve the time-consuming and costly search for new antibiotics as bacteria and other microorganisms continue to evolve resistance to the current array of drugs.
The new model, called SyntheMol-RL, is trained to explore a vast chemical space of up to 46 billion possible compounds. This is far beyond what can be realistically tested in a laboratory, where even large screens reach about 1 million molecules. Utilizing a set of approximately 150,000 molecular “building blocks” and 50 chemical synthesis reactions, the AI model is designed to generate structurally novel antibiotic candidates.
“In the lab, we can build compounds using a series of small chemical fragments that can be glued together like molecular Lego blocks,” says Assistant Professor John Stokes, whose lab developed the new model. “SyntheMol-RL composes these fragments in different ways, faster than humans could ever do, and based on that knowledge creates new, larger compounds that should have antibacterial properties.”
members of stokes Michael G. DeGroot Institute for Infectious Diseasessays that although generative AI is becoming increasingly effective in designing novel antibiotic candidates, it remains difficult to assess the key properties that determine the clinical viability of potential drugs without large-scale and expensive clinical testing.
“Even if a new chemical is found in the lab that has antibacterial properties, it’s useless if it doesn’t dissolve in the body, is toxic to human cells, or can’t be metabolized and excreted after it’s done its job,” he explains. “Bleach is antibacterial, and so is flame, but they clearly don’t check the other boxes. Good drug candidates have to meet several different criteria, otherwise they’ll never become actual drugs.”
Past iterations of SyntheMol We independently designed molecules with antimicrobial activity without considering these other important properties. But over the past two years, Stokes’ team has worked with collaborators at Stanford University to refine the model so that it only produces antimicrobial compounds that are easy to develop in the lab and likely to dissolve in the body.
“There are a lot of contradictions between antimicrobial compounds and water-soluble compounds,” says Gary Liu, a graduate student in Stokes’ lab and lead developer of the new model. “Previous research has shown that by filtering out antibacterial compounds, and soluble rear Our prompts often resulted in significantly fewer viable drug candidates. Therefore, we incorporated solubility into the production process, and this model now allows us to efficiently design more clinically promising antibiotic candidates. ”
in new researchwas published on April 23rd and was selected as the cover of the June issue. molecular systems biologyStokes’ team tested a reinforcement model. They were tasked with producing water-soluble antibiotics that could treat infections caused by bacteria. Staphylococcus aureus This was colloquially known as a “staph infection” and quickly became a hit with some.
Among a batch of 79 antibiotics proposed as models, Stokes’ group found one compound of particular interest. It is a novel water-soluble compound that appears to have antibiotic activity against antibiotics. Staphylococcus aureus.
The new computer-designed drug candidate, called Synthecin, was formulated as a topical cream in the lab and tested on drug-resistant wound infections in mouse models.
“Synthecin was very effective at controlling the infection,” says Dennis Kataktan, a graduate student in Stokes’ lab who led the wet lab portion of the study. “It works very well as a topical drug and shows early promise as something that could be applied or optimized for systemic use in the future.”
The new study highlights the promise of synthesin, but the research team has not yet determined how the drug inhibits bacteria. This is an important step in determining its safety profile and, therefore, its potential to someday be introduced into the clinic, Stokes said. His group is currently actively researching these important “mechanisms of action.”
But regardless of how these studies pan out, the research group sees Synthesin’s discovery as proof that their AI models can rapidly generate high-potential drug candidates, shifting the burden of drug discovery from finding viable compounds to designing and optimizing them.
Stokes says this change is important not just for antibiotic discovery, but for all areas of biochemistry.
“We used the model to design new antibiotics, but much more is possible,” says Stokes, a faculty member at the institute. Marnix E. Hersink School of Biomedical Innovation and Entrepreneurship and its executive members nexus health. “We built this product to be disease agnostic, which means we can easily generate new drug candidates for diabetes, cancer, and other indications.”
Stokes’ lab continues to enhance SyntheMol, and a more robust version will be available later this year.
