Drug development is often a long and tedious process. In some cases, it can take more than a decade and billions of dollars for a treatment to reach patients. In fact, more than 90% of drug trials fail in the early stages.
New research from Queen’s University aims to change this by using artificial intelligence (AI) to determine successful drug candidates and has been awarded the 2025 John Charles Polanyi Prize in Chemistry.
The award is presented annually to five researchers who are early in their careers, are pursuing postdoctoral research, or have been appointed to a faculty position at an Ontario university.
The 2025 prizes are worth $20,000 each and are broadly defined in the categories of physics, chemistry, physiology or medicine, literature, and economic science.
In an interview with journalDr. Fawang Meng, winner of the 2025 Polanyi Prize in Chemistry and postdoctoral fellow in the Banting Department of Chemistry, shared insights about his research.
“My research focuses on the early stages of drug discovery, and I hope to identify molecules and compounds with desirable properties that make them good candidates for clinical trials,” Meng said.
Meng’s research uses machine learning to help identify compounds that meet multiple properties, such as efficacy, safety, and chemical stability, before proceeding with drug development.
According to Meng, datasets generated by advances in computing power and automated experimentation systems are key to training modern machine learning models. “We have amassed a huge amount of data points, and those data points provide the foundation for building really powerful machine learning models,” Meng added.
By analyzing these datasets, AI learning models can predict how different compounds will behave. “Based on those predictions, we can select some of the most promising ones. [compounds]and send those compounds to the experimental setup. New data can also be fed back into the model, so the model can use it for the next round of predictions,” Meng said.
Before the widespread use of machine learning, computational drug discovery relied on traditional simulation methods, which were not always reliable.
“Speed can be a very big issue. In some cases, it’s so slow that it’s difficult to tell whether the predictions are accurate or not,” Meng pointed out. “What sets AI models apart is their efficiency, speed, and accuracy.”
However, applying machine learning to chemistry presents unique challenges regarding the quality of available data. “Data sets can have a lot of missing values,” Meng says. “The other thing is that datasets can be unbalanced, and some datasets are actually quite noisy because of their experimental settings.”
“Discarding data with missing values is normal and often leads to information loss. What I did was try to build a model that maximizes the value of each data point,” Meng added.
Improving how models handle incomplete data sets could lead to more reliable predictions and ultimately better drug candidates.
For now, Meng sees her work as part of a broader effort to enhance the tools scientists use to discover drugs.
tag
Department of Chemistry, Drug Trials, Machine Learning, Polanyi Prize
All final editorial decisions are made by the Editor-in-Chief and/or Editor-in-Chief. Under no circumstances should you contact the author, be targeted, or be harassed. If you have any complaints about this article, please send your comments to journal_editors@ams.queensu.ca.
