���ϳԹ���Դ

Developing a new medicine can take more than a decade of research and billions of dollars in investment, yet many promising drug candidates do not make it to patients, with problems surfacing often in late-stage testing.

���ϳԹ���Դ researcher Dr. Fanwang Meng (Department of Chemistry), a Banting Postdoctoral Fellow, is working to address this costly challenge. He is designing machine learning systems that aim to detect those potential problems much earlier, and it has earned him the 2025 John Charles Polanyi Prize in Chemistry, one of Ontario’s top honours for early-career scientists.
 

Catching problems early in drug development

Computational drug discovery depends on experimental data, but that data is often limited, noisy, or incomplete. Researchers may test hundreds or thousands of molecules before identifying one that appears both safe and effective. Even then, molecular property problems can arise at a later stage and halt further development, preventing the candidate molecules from getting into clinical usage.

Dr. Meng builds algorithms that analyze patterns in chemical structure and experimental data to predict which compounds are more likely to behave safely and effectively in the human body. By identifying similarities in structure and behaviour, the system helps researchers decide which molecules are selected for further testing.

“Drug discovery is full of uncertainty with nearly 94 per cent of drug candidates failing before reaching patients because problems arise that we cannot easily see at the start,” says Dr. Meng. “By designing smarter machine learning models, we can predict risks earlier and uncover promising treatments that might otherwise be missed.”
 

Strengthening models against imperfect data

A persistent challenge in biomedical research is data quality, including missing values, imbalanced or biased data structures, and data noise. This accounts for many practical factors, such as some compounds not being available for biological testing, leading to missing values, and publication bias favouring only positive results, leading to imbalanced or biased data. These factors weaken predictions.

To address this, Dr. Meng trains his models to work effectively even when data is imperfect. He evaluates them using highly biased benchmarks, including the blood-brain barrier permeability dataset developed during his PhD research. This testing helps ensure the models remain reliable under real-world research conditions. More dependable models can support a more efficient development process, including work on treatments for diseases such as malaria.
 

Back-to-back Polanyi recognition at ���ϳԹ���Դ

Dr. Meng’s recognition follows last year’s Polanyi Prize in Chemistry awarded to ���ϳԹ���Դ researcher Rachel Baker. Since 2007, five researchers in the Department of Chemistry have received the honour, including Dr. Meng and his postdoctoral supervisor, Dr. Farnaz Heidar-Zadeh, reflecting the department’s long-standing strength in chemical research.

“Congratulations to Dr. Fanwang Meng on receiving the Polanyi Prize for Chemistry,” says Principal and Vice-Chancellor Patrick Deane. “His research is reshaping the field of drug discovery, accelerating the identification of new medicines, and contributing to the development of safer, more effective treatments for Ontarians. Through the open-source release of his models and datasets, he is enabling collaboration across the global research community and exemplifying ���ϳԹ���Դ’s commitment to innovation in service of society.”

Learn more about