Using AI to Create More Precise and Personalized Drug Delivery Solutions

Daniel Reker received two research awards from the National Institutes of Health: the National Institute of Biomedical Imaging and Bioengineering (NIBIB) Trailblazer Award and a National Institute of General Medical Sciences (NIGMS) Maximizing Investigators’ Research Award (MIRA).

An assistant professor of biomedical engineering at Duke University, Reker will use these awards to design machine learning models that can improve the development of more effective drug delivery tools.

The field of drug delivery is centered on the creation of technologies and formulations that can ensure that therapeutic drugs reach their intended target. Without these delivery mechanisms, many lifesaving therapies would be quickly removed from the body before they had a chance to take effect. Delivery vehicles can also help limit the side effects of toxic drugs, like chemotherapies, by ensuring that the drugs reach the intended tissues without causing wider damage to the patient.

“Most of the drug delivery development follows a trial-and-error approach, which is laborious and costly, or relies on a one-size-fits-all approach, where researchers will broadly apply one delivery vehicle to lots of different drugs,” explains Reker. “Although the drugs still work, it often results in suboptimal drug delivery with potentially lower efficacy and additional side effects.”

With work that spans computational biology, chemistry, engineering and pharmacology, Reker specializes in using machine learning approaches to design novel drug delivery vehicles that can more effectively deliver drugs. Now, Reker and his team will use these grants to build on this work to develop new machine-learning algorithms that can design the best drug delivery vehicle for specific therapies, even personalizing them to the needs of different patients.

Reker will use the funding from the NIBIB Trailblazer Award, which will provide his team with $580,000 over three years, to design nanoparticles that can precisely target diseased tissue or cancerous cells.

“Nanoparticles can safely deliver therapeutics and vaccines throughout the body, but they are difficult to make and can only carry a small amount of a drug,” says Reker. “Only about 30 nanoparticles have been approved by the FDA, and none of the approved formulations can precisely target specific tissues or cells yet.”

Recognizing these limitations, Reker previously used machine learning to identify, design and optimize 100 brand-new nanoparticles. The team will use a similar approach to create models that attach antibodies to new nanoparticles. By pairing the nanoparticle formulations with the right antibodies, the nanoparticles could precisely target diseased tissue or cancer cells, rather than broadly circulate throughout the body.

Reker will pursue related research with the support of the MIRA, which will provide more than $1.8 million in funding over five years. Given to NIH Outstanding Investigators, the MIRA funding mechanism was created to provide researchers with greater stability and flexibility while they pursue ambitious scientific projects.

While Reker will also use the MIRA funding to develop machine learning algorithms for targeted nanoparticle formulations with other targeting functionalities like peptides, he’ll also pursue two new projects.

The first project will involve identifying excipients to modulate how the microbiome metabolizes drugs. Most drug products, like pills, gels and capsules, contain both an active pharmaceutical ingredient and a specific mixture of ‘inactive’ ingredients, called excipients. While some ingredients are typically added to alter the physical properties of a medication, like its appearance, researchers can also add functional excipients, which can change how a drug is transported and metabolized across tissues to improve the drug’s uptake.

“For example, some chemotherapeutics can cause painful gastrointestinal side-effects because they are transformed into toxic molecules by the body’s microbiome, but if we could add a functional excipient that changes the metabolism of the microbiome, we could theoretically prevent those toxic molecules from being created and make the medications safer,” says Reker.

But to identify and test every excipient would take billions of experiments. Instead, the team will train machine learning models to predict the outcomes of those experiments and prioritize the most promising targets.

For the final project, Reker will develop models to improve the design of prodrugs, a specific inactive version of a medication that is chemically converted into an active drug as it’s metabolized by the body. Prodrugs are much easier to store, transport and administer than many other therapeutics, as the inactive versions are often more stable at room temperature and can be taken orally.

“Despite their benefits, prodrugs only make up about 10 percent of all approved drugs because they are quite difficult to design,” says Reker. “We’re hoping our lab can help harness the full potential of these therapies by using machine learning to design new prodrugs that could help them become safer, more effective and more widely available.”

Although the MIRA is primarily focused on research, the program also encourages faculty to serve as mentors to newer researchers. For Reker, this work is second nature.

“I consider my primary role at Duke to be a mentor and a sponsor of trainees and students, and these awards help me create projects that train students in a unique combination of skills that will make them future leaders in their chosen career paths,” he says. “By knowing how to speak the language of drug delivery and the language of machine learning, the next generation of engineers will be able to shape the future of medicine.”