Artificial intelligence (AI) has become extremely popular. It powers models that help us drive vehicles, proofread emails, and even design new drug molecules. But just like a human, it is difficult to read the minds of AI. Explainable AI (XAI), a subset of the technology, could help us do this by justifying a model’s decisions. And now researchers are using XAI not only to examine predictive AI models more closely, but also to look deeper into the field of chemistry.
The researchers will present their results at the fall meeting of the American Chemical Society (ACS).
The sheer number of applications of AI has made AI nearly ubiquitous in today’s technological landscape. However, many AI models are black boxes, meaning it is not clear exactly what steps need to be taken to achieve a result. And if that result is anything like a potential drug molecule, not understanding the steps could raise skepticism among scientists and the public alike. “As scientists, we like justification,” explains Rebecca Davis, a professor of chemistry at the University of Manitoba. “If we can come up with models that can provide some insight into how AI makes its decisions, this could potentially make scientists more comfortable with these methodologies.”
One way to provide that justification is with XAI. These machine learning algorithms can help us see behind the scenes of AI decision-making. While XAI can be applied in a variety of contexts, Davis’ research focuses on its application to AI models for drug discovery, such as those used to predict new antibiotic candidates. Given that thousands of candidate molecules can be screened and rejected to approve just one new drug – and antibiotic resistance poses an ongoing threat to the efficacy of existing drugs – accurate and efficient prediction models are critical. “I want to use XAI to better understand what information we need to teach chemistry to computers,” said Hunter Sturm, a graduate student in chemistry in Davis’ lab who is presenting the work at the meeting.
The researchers started their work by feeding databases of known drug molecules into an AI model that would predict whether a compound would have a biological effect. They then used an XAI model developed by collaborator Pascal Friederich of Germany’s Karlsruhe Institute of Technology to investigate the specific parts of the drug molecules that led to the model’s prediction. This helped explain why a particular molecule did or did not have activity according to the model, and that helped Davis and Sturm understand what an AI model might consider important and how it creates categories after examining many different compounds.
The researchers realized that XAI can see things that humans might have missed; it can consider many more variables and data points simultaneously than a human brain. For example, when screening a series of penicillin molecules, the XAI found something interesting. “Many chemists consider the penicillin nucleus to be the critical site for antibiotic activity,” says Davis. “But that’s not what the XAI saw.” Instead, it identified structures attached to that core as the critical factor in the classification, and not the core itself. “This could be why some penicillin derivatives with that core show poor biological activity,” Davis explains.
In addition to identifying important molecular structures, the researchers hope to use XAI to improve predictive AI models. “XAI shows us what computer algorithms define as important for antibiotic activity,” Sturm explains. “We can then use this information to train an AI model on what it should be looking for,” Davis adds.
Next, the team will work with a microbiology laboratory to synthesize and test some of the compounds that the improved AI models predict would act as antibiotics. Ultimately, they hope that XAI will help chemists create better, or perhaps entirely different, antibiotic compounds that can help stem the tide of antibiotic-resistant pathogens.
“AI causes a lot of distrust and uncertainty among people. But if we can ask AI to explain what it is doing, this technology is more likely to be accepted,” says Davis.
Sturm adds that he thinks AI applications in chemistry and drug discovery represent the future of the field. ‘Someone has to lay the foundation. That’s what I hope I do.’
The research was funded by the University of Manitoba, the Canadian Institutes of Health Research and the Digital Research Alliance of Canada.
