Jan. 6, 2026

(Shutterstock)

A recent publication in ACS Pharmacology & Translational Science shares how a team of NIH researchers developed an innovative approach for streamlining the process of probe discovery for a family of enzymes implicated in a wide range of diseases, from cancer to neurodegenerative disorders to metabolic disease.

Chemical probes are small molecule tools essential for exploring the roles of proteins and enzymes in biological systems, diseases, and medical interventions. Probes that display enough drug-like properties may even be developed into therapeutics. Identifying and developing probe candidates involves large investments of time, labor and resources, so innovative new approaches for expediting the process are in high demand.

This is especially true for a historically challenging family of protein targets known as aldehyde dehydrogenases (ALDHs). ALDHs are responsible for the breakdown of toxic aldehydes, a class of chemicals generated from a variety of metabolic reactions that occur throughout the human body. Excess accumulation of aldehydes can lead to DNA damage or even cell death. To prevent this, each of the 19 unique ALDH enzymes is tailored to catalyze the conversion of aldehydes into less harmful products. For example, ALDH2 breaks down the toxic acetaldehyde generated from the metabolism of ethanol consumed in alcoholic beverages. In the context of disease, elevated ALDH levels are hallmarks of several cancers and are strongly linked to treatment resistance, poor prognosis, and aggressive tumor growth. ALDHs also play a role in obesity and some rare diseases. Understanding how specific ALDH enzymes are involved in disease is critical for determining which members may be viable targets for drug discovery and development.

A team of NIH scientists led by Natalia Martinez, Ph.D., and Alexey Zakharov, Ph.D., has created a new strategy that combines state-of-the-art high-throughput methodologies with machine learning (ML) to screen large chemical libraries and identify probe candidates. As an ambitious test case for the team’s novel approach, the goal was to scale the strategy to uncover multiple probes against the ALDH family simultaneously. Dr. Martinez explained, “Some cancers have very high amounts of one ALDH compared to healthy cells. Other cancer cells will show increased amounts of a completely different ALDH protein or combination of proteins. It’s not one solution that fits all.”

The NIH team virtually screened an internal collection of 174,000 compounds against four ALDH proteins. “By combining traditional modeling with artificial intelligence methods, we screened a much larger collection of compounds. With this hybrid approach, we identified chemical probe candidates for four targets at the same time, which exponentially decreases the time and resources used,” said Sankalp Jain, Ph.D., a co-lead author on the publication. Dr. Martinez emphasized that the chemical probe candidates identified for the four ALDH proteins are selective, meaning they were not found to target other ALDH proteins, at least in the panel tested. This finding is an encouraging sign that these compounds might be able to prevent harmful side effects caused by widespread ALDH inhibition.

The data generated from this study are publicly available, allowing other researchers to advance their research efforts — or avoid ill-fated endeavors. Adam Yasgar, another co-lead author on the publication, stressed, “Chemical probes are crucial for validating a target, but they are equally valuable for ruling one out. We are helping researchers avoid dead ends and saving significant time and resources by making this data public.”

For the NIH team, next steps include screening larger virtual compound collections and extending this approach to additional ALDH isoforms. The researchers also will work toward refining lead compounds for potency, selectivity and developability, as well as pursuing in vivo proof-of-principle studies.