Resetting the Clock on Discovery

Article Content

This story is from the 2026 issue of Discoveries, a UC San Diego Health Sciences magazine.

Not long ago, discovering a new drug followed a familiar, painstaking arc. Scientists would spend years identifying a biological target, screening thousands of compounds and refining promising leads in the lab — often with no guarantee that this effort would ultimately benefit patients. Today, that timeline is being dramatically rewritten.

At UC San Diego Skaggs School of Pharmacy and Pharmaceutical Sciences, artificial intelligence (AI) and advanced computational models are transforming how medicines are discovered. Interdisciplinary teams of chemists, biologists, computer scientists and data experts are using machine learning, high-performance computing and physics-based modeling to compress years of discovery into months — or sometimes seconds.

“We’re not just speeding things up,” said Distinguished Professor William Gerwick, PhD ‘81, a trailblazer in AI-enabled marine drug discovery. “We’re asking questions that simply couldn’t be asked before.”

Mining the ocean with machine learning

Gerwick’s laboratory begins drug discovery far from a computer screen. Armed with scuba gear, he and his students dive in tropical waters — from Panama to Papua New Guinea — collecting marine algae and cyanobacteria.

Gerwick has a joint appointment as professor of marine chemistry and geochemistry at the Scripps Institution of Oceanography, and the organisms he studies thrive in competitive underwater environments where they produce an astonishing diversity of chemical compounds — many with potent biological activity.

Historically, identifying those compounds was a bottleneck. Once an extract showed promise — killing cancer cells or parasites, for example — scientists could spend weeks or months deciphering its molecular structure using experimental techniques, such as nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry.

That changed when Gerwick teamed up with computer scientists to develop AI tools that learn how molecules “look” in NMR data. One system, called Small Molecule Accurate Recognition Technology (SMART), uses deep learning to rapidly generate structural hypotheses for unknown compounds.

In one early test, a postdoctoral researcher fed NMR data from a partially purified compound into SMART 2.0. Eight seconds later, the system returned a ranked list of likely molecular structures.

“That was an aha moment,” Gerwick recalled. “What used to take weeks suddenly took seconds.”

Despite its efficiency, scientists such as Gerwick emphasize that AI doesn’t replace human expertise; it proposes hypotheses. Chemists still confirm structures, test biological activity and refine molecules. But by removing a major roadblock, AI allows researchers to move faster toward understanding how a compound works and whether it can be developed into a medicine.

Gerwick’s team is now pushing this concept further, developing AI tools that can predict whether a newly discovered molecule might have activity against cancer or other diseases based solely on its chemical structure.

Zooming in on proteins

While Gerwick looks for small molecules in nature, Skaggs School of Pharmacy Professor Irina Kufareva, PhD, looks at the larger proteins driving our physiology from within.

Proteins control nearly every process in the body, but for decades scientists knew the structures of only a small fraction of them. Predicting how a protein folds — or how it changes shape when interacting with a drug — was one of biology’s grand challenges.

That all changed in the early 2020s with the arrival of AI systems that predict protein structure, such as the Nobel Prize-winning AlphaFold.

“Suddenly, we could predict protein structures with unprecedented accuracy,” Kufareva said. “It completely changed the field.”

These breakthroughs are particularly important for G protein-coupled receptors (GPCRs), a massive family of cell-surface proteins involved in pain, inflammation, metabolism and cancer. Roughly one-third of all approved drugs target GPCRs, yet many of these proteins remain poorly understood.

Using AI-predicted structures, Kufareva’s lab explores how drugs bind to these receptors and why certain mutations cause disease, which may help scientists design improved therapies for human diseases. In recent work, her team used AI-enabled modeling to design an enhanced version of a natural signaling molecule involved in inflammatory bowel disease, creating a potential starting point for future therapeutics.

However, Kufareva also cautioned that AI is not magic. “You still need physics. You still need chemistry,” she said. “The most powerful approaches combine machine learning with fundamental scientific principles."

This is a philosophy that resonates across Skaggs School of Pharmacy computational drug discovery efforts.

Tomorrow’s drugs arriving at an accelerated speed

For Skaggs School of Pharmacy Assistant Professor Adrian Jinich, PhD, AI offers a way to tackle some of the world’s deadliest diseases, including tuberculosis and malaria. 

Jinich’s lab blends machine learning with experimental biochemistry to predict how pathogens survive. One of his most striking projects began with an idea by renowned expert on malaria Skaggs School of Pharmacy Associate Dean for Research and Innovation Elizabeth Winzeler, PhD — and a very young student.

A high school junior with a passion for machine learning joined Jinich’s lab and began designing proteins on a computer that could bind to a key malaria parasite target. Using UC San Diego’s Supercomputer Center and AI-driven protein design tools — including AlphaFold — the student screened hundreds of thousands of candidates.

The results astonished the team. Of eight proteins synthesized and tested in the lab, three strongly inhibited malaria infection. Those proteins were then delivered to Winzeler for further research.

“That simply wouldn’t have been possible a few years ago,” Jinich said. “AI lets small teams — and even students — do work that once required enormous resources.”

Jinich is now applying similar approaches to tuberculosis, using machine learning to narrow millions of possibilities down to a manageable set of candidates for experimental testing.

Data research as the fuel for discovery

AI thrives on data, and Skaggs School of Pharmacy Professor Michael Gilson, MD, PhD, has spent decades building one of the field’s most important data resources: BindingDB.

BindingDB is an open, National Institutes of Health (NIH)-funded database containing millions of experimentally measured interactions between proteins and small molecules. Today, it is one of the most widely used datasets for training AI models in drug discovery.

“AI doesn’t work without high-quality data,” said Gilson, who is co-director of the UCSD Center for Drug Discovery Innovation. “BindingDB helps provide that ground truth.”

Beyond data curation, Gilson collaborates closely with computer scientists, including Rose Yu, PhD, associate professor in the Department of Computer Science and Engineering, to develop new AI methods that respect the laws of physics and chemistry. Their joint work focuses on generative models that can propose entirely new drug-like molecules — balancing novelty with the practical reality that those molecules must be synthesizable in the lab.

This kind of cross-disciplinary collaboration is a hallmark of UC San Diego, where faculty routinely bridge health sciences, engineering and computer science to solve complex problems.

Building the future, one student at a time

For all the power of artificial intelligence, one theme emerges across UC San Diego’s AI-based drug discovery research: People still matter most.

AI systems generate hypotheses, not answers. They accelerate discovery, but they do not replace scientific judgment.

That’s why training the next generation of scientists is central to the Skaggs School of Pharmacy mission. For example, Jinich teaches a course in the Department of Chemistry and Biochemistry on applying AI to chemistry and biochemistry, where undergraduate and graduate students learn to use AI tools responsibly and creatively. As part of the course, students design projects, write code and learn to distinguish reliable predictions from “robotic fantasy.”

This commitment to training extends into the lab as well. Kufareva mentors dozens of undergraduates in her lab, emphasizing critical thinking in an era when answers seem instantly available. Gerwick also stresses the importance of teaching trainees the fundamentals of chemistry so they can evaluate AI-generated results with expert eyes.

“We still need natural intelligence,” Kufareva said. “AI is a powerful assistant — but only if we train people who understand what it’s doing.”

As AI continues to reshape drug discovery, Skaggs School of Pharmacy stands at the forefront not just because of its algorithms or computing power but because of its people. By combining advanced technology with deep scientific expertise and a commitment to education, the school is not only accelerating the search for new medicines; it is redefining how that search is done.

And in doing so, it is preparing a generation of scientists ready to carry medicine into a faster, smarter future.