Two researchers from the U of M, Rebecca Davis, an associate professor in the department of chemistry, and Hunter Sturm, a PhD candidate, are pioneering a new approach to antibiotic drug development using artificial intelligence (AI).
Davis leads the U of M’s Davis Research Group, located out of the Parker Building on the Fort Garry campus.
The group utilizes computational and physical organic chemistry to develop predictive models that address problems pertaining to synthetic method development. Also, the group hopes to contribute to efforts in antibiotic discovery.
Davis and Sturm’s research focuses on how Explainable AI (XAI) can be applied in AI models for antibiotic drug development.
XAI is a subset of AI that helps to justify the reasons behind a machine learning algorithm’s decisions and predictions.
Unlike traditional AI, this AI makes complex AI decisions more transparent by offering clear insight into how it arrived at its predictions.
Sturm emphasized the importance of transparency in AI, stating that they intend to teach the AI models to show the steps they take to arrive at a conclusion, rather than just assuming the steps they took are accurate.
He stated that, “we know that AI is only as good as the information that is fed into it, so we want to remove the mystery and train the AI model on what it should look for and understand the steps it has taken.”
Davis also clarified that it is not their first time working with this AI model.
She mentioned that they have previously used this model to classify molecule aggregators — a type of false positive in activity screens — and that success is their main inspiration to believe that XAI can also be useful in the development of drugs.
According to Davis, they will be “using both deep learning models and XAI to identify molecular scaffolds that can be druggable for this research.” Their goal in this research is to lower the cost and the time it takes to bring drugs to market.
Sturm said, “we are going to be able to speed up the drug discovery process so finding new antibiotics will happen at a faster rate, meaning that we would have more antibiotics in the market in a faster time frame.”
The team believes that XAI would promote a safer and more ethical application of AI in drug development.
Davis expressed that the XAI’s interpretation of the explanations can improve the model’s validation and refinement, as it allows them to evaluate if the AI predictions align with experimental known mechanisms of the drug.
In addition, this project is a part of a broader international collaboration that includes experts from various fields, such as microbiology, bioinformatics and computer science.
This multidisciplinary team is dedicated to refining the antibiotic discovery pipeline, which could lead to more effective and timely drug development.
Davis and Sturm also presented their work at the fall meeting of the American Chemical Society, as they are among a small group of researchers in North America who are trying to inform the science community around XAI and encourage an open mind toward new methodologies.