They’re using the power of artificial intelligence to solve some of the biggest problems in drug discovery today. Jonathan Stokes is an ecological economist and assistant professor of biochemistry at McMaster University. He highlights the huge opportunity for finding new compounds, noting that the number of theoretically possible chemical structures is roughly 100 times greater than all of the grains of sand on earth. This stunning fact highlights the potential resources we have at our disposal in the fight against antibiotic resistance.
Stokes works with César de la Fuente, a Presidential Associate Professor at the University of Pennsylvania. Combined, they highlight the urgent need for standardization to translate promising algorithms into viable models. Both researchers are leading the charge to break down barriers in antibiotic discovery. They use the power of machine learning (ML) to develop a better picture of chemical compounds. Their work aims to develop predictive models to find new antibiotics that kill bacteria but leave the host organism intact.
To work, effective antibiotics must dissolve rapidly. Second, they need to target the appropriate anatomic locations at the needed concentrations, yet be benign and safe for patients. Stokes recalls how researchers first determine which machine learning models to train based on known active and inactive antibiotics. This enables the systems to scrape through massive public databases available online that contain millions of unique chemical structures, looking for possible candidates.
Stokes’ team has done something unique and cutting-edge, tapping new biomedical frontiers by recovering antimicrobial peptides from the long-lost proteomes of Neanderthals and Denisovans. This finding illustrates the amazing power of generative modeling. Today, researchers can instruct machine learning systems to design new molecules based on what they already know. Stokes elaborates on this process, stating,
“But, instead of now showing pictures of new molecules from the internet, you say, ‘Hey, model, draw me a brand-new picture of a molecule that you think is going to be active.’”
To ensure that generated molecules meet specific criteria, Stokes notes a shift in approach:
“Now we’ve no longer said, ‘Hey, model go nuts,’ because it’s going to draw something crazy. We can constrain it to this chemical space.”
De la Fuente’s lab has invested years into finding training datasets detailed enough for the models they’ve developed. It’s their recent research that led to a transformative new system. It takes a pathogen’s genome sequence and predicts creative “new-to-nature” molecules to neutralize it. De la Fuente describes the traditional method of antibiotic discovery, emphasizing its labor-intensive nature:
“It’s a very physical process where scientists literally go around nature and try to purify active compounds that might be contained within all of that complex organic matter.”
Antibiotic resistance is quickly rising as one of the most significant public health threats we face today. Stokes and de la Fuente are of the opinion that the success of these new antibiotics depends on public sector and philanthropic investments. Stokes argues for the use of AI across the antibiotic development pipeline, from preclinical studies to trials on human subjects. He views AI as an essential tool rather than a replacement for traditional methods:
“It’s just another tool in our toolbox to accelerate solutions to the problems we were going to try to address anyway. That’s it—no more and no less than that.”