Research for the Holy Grail

Gabe Kwong imagines a world where a delicious cup of enhanced yogurt followed by a routine blood test at the local clinic could signal the presence of cancer long before symptoms arise.

It isn’t a far-fetched fantasy, either, thanks to investigators like Kwong and his colleagues at Winship Cancer Institute of Emory University, Georgia Tech and beyond. Their research in a diverse range of disciplines—including biomedical engineering, immunology, computation and artificial intelligence (AI)—is edging us toward a future in which clinicians can screen for multiple cancers earlier than ever, leading to better treatment outcomes, longer lives and a more robust health care system.

While companies raise billions of dollars to advance cancer detection blood tests, Kwong and fellow Winship researcher Philip Santangelo are leading ongoing multimillion-dollar federal efforts, launched by the Biden Administration, to drastically reduce the cancer mortality rate.

“It’s challenging to detect cancer at its earliest stages, but that is the objective,” says Kwong, a member of Winship’s Cancer Immunology Research Program and an assistant professor in the Wallace H. Coulter Department of Biomedical Engineering at Emory and Georgia Tech. “That’s the dream, and it’s within reach.”

HAYSTACK NEEDLE

Finding the earliest traces of cancer in the blood is absurdly difficult, like searching for the proverbial needle in a haystack. Although tumors shed mutated DNA and enzymes into the bloodstream, it’s in tiny, almost invisible amounts. “That’s the core challenge,” says Kwong, who leads the Laboratory for Synthetic Immunity, where his team engineers immune-based medicines to intercept and treat disease. This includes the development of synthetic antigen “flags” that train immune cells to recognize tumors, and ultra-sensitive sensors designed to quickly detect cancers—which is about as easy as hearing a whisper in a stadium of people. “The most promising approaches so far have been focused on sequencing circulating tumor DNA,” Kwong explains. “The challenge is sensitivity—being able to pick up one molecule of DNA in a 10-milliliter blood draw.”

Even the most advanced companies in this space have struggled. Silicon Valley biotech firm GRAIL completed a first trial of its Galleri test in 15,000 patients with stage one cancer. The test had a sensitivity level of just 17%.

“That’s the sobering reality,” says Kwong, whose own lab is taking this approach: Instead of just listening for stray tumor DNA, he wants to amplify the signal.

So, Kwong and his team are building synthetic biosensors that act like tracers, an approach that has been guided by advances in imaging technology, “where you give a patient a contrast agent that reacts in some way to produce a disease contrast,” he says. “Our sensors use tumor proteases [enzymes] to turn on the signal, creating contrast between healthy and diseased tissue.”

They use nanoparticles adorned with special peptides, which respond to enzymes that are secreted by tumors. After they’re injected, the nanoparticles circulate in the body until encountering those enzymes, which cleave the peptides (essential building blocks of proteins), releasing detectable fragments in blood or urine. “Our sensors use tumor proteases that turn on the signal,” Kwong says. “And that creates the contrast between healthy and diseased tissue.” The system proactively coaxes tumors into revealing themselves, rather than passively waiting for them to shed DNA.

THE MOONSHOT

Kwong and fellow biomedical engineering researcher Santangelo are leading ambitious initiatives supported by the Advanced Research Projects Agency for Health, or ARPA-H. Modeled on DARPA, the defense research agency that helped create the Internet, ARPA-H aims to cut the cancer death rate in half within 25 years.

Santangelo’s $24 million piece of former President Biden’s “cancer moonshot” isn’t focused on detection but on treatment. His lab develops RNA-based therapeutics—think of cancer vaccines. “These are cancer therapeutics that act like vaccines,” says Santangelo, a professor in the Coulter Department of Biomedical Engineering and a member of Winship’s Cancer Immunology Research Program. “They elicit an immune response or manipulate genes in immune cells to enhance tumor killing.”

While his immediate goal is therapy, Santangelo sees a bridge to early detection. “Part of our goal is to understand the immune response in patients and, ultimately, use RNA probes to detect pre-cancerous tissue changes,” he says. For now, though, his work sits squarely in treatment: reducing the tumor burden, extending lives, laying the groundwork for personalized cancer immunology. “If we keep trying to do the same old thing over and over again, we’re not going to get better results,” he says. “We need new platforms.”

For Kwong, ARPA-H is like a dream accelerator—his biosensor program is at least 10 years old, and he’s described the federal program as a mechanism to implement his vision, “and go at light speed.” He adds, “That's pretty exciting.” He’s leading a $50 million effort with a team of collaborators from multiple institutions, which includes the development of some amped up, high tech yogurt.

Kwong’s ARPA-H team includes Tal Danino, a researcher at Columbia University who engineers probiotics as living sensors. It’s the kind of work that may lead to Star Trek medicine. “What we found is that these probiotics can live in the center of a tumor, going undetected by the body’s immune system, basically colonizing the center of tumors,” Kwong says. “So, imagine these safe, engineered probiotics that don’t survive very long inside the body.”

The idea is, a patient would eat the yogurt then wait 24 to 48 hours before having his blood or urine tested. The probiotics would be designed to release a signal if early-stage cancer or pre-cancerous evidence is detected. “In patients without tumors, there would be no signal,” Kwong says.

COMPUTATIONAL INTERPRETATION

Even the best, most accurate biosensors are useless without someone or some way to interpret their signals. That’s where Peng Qiu, one of Kwong’s ARPA-H collaborators, comes in.

Qiu, an associate professor in the Coulter Department of Biomedical Engineering and a member of Winship’s Discovery and Developmental Therapeutics Research Program, is a computational biologist whose lab leverages its expertise in machine learning and bioinformatics to make biomedical discoveries. His job is to basically turn raw experimental data into reliable insights. “We combine single-cell data, high-throughput experimental screening and computational analysis to identify synthetic biomarkers for early cancer detection,” he explains.

If you think of Kwong as the conductor directing the instruments, then Qiu is the composer, writing the sheet music. His models show how sensors behave, cross-check the results and integrate complex datasets into patterns that can guide clinical decisions, which are made by human beings.

“AI may be powerful for making some specific predictions, like cancer recurrence risk,” Qiu says. “But progress in biomedicine comes in small, incremental steps made by people.”

He stresses the human side of collaboration. “Scientific projects take a village—experimentalists, computational researchers and people to communicate with the broader community,” says Qiu, who is cautious, pragmatic and realistic, but nonetheless calls early detection, “one of the potential holy grails out there.”

He adds, “A simple blood screen can have a huge impact, and that fuels my passion for this work. Because if we can catch cancer early, then a more positive patient outcome is almost always guaranteed.”

---

---

THE AI OPPORTUNITY

Anant Madabhushi, executive director of the Emory Empathetic AI for Health Institute (AI Health) and a member of Winship’s Cancer Immunology Research Program, is using AI to search for cancer clues in data that already exist. Instead of designing new sensors, he builds algorithms to dig up routine clinical information—like imaging scans or even photographs of the eye.

Recently, Madabhushi led a multi-institutional effort to develop an AI system called RetHemo, which can analyze retinal scans and spot subtle changes in blood vessels. These early warning signs tied to chronic inflammation often precede blood cancers like leukemia, myeloma and lymphoma.

In a study of more than 1,200 patients, RetHemo predicted the future development of cancer—up to 10 years before diagnosis—and performed better than traditional clinical predictors. “A simple eye photo, meant to screen for glaucoma, can hint at your cancer risk long before the disease takes hold,” says Madabhushi. “This is opportunistic AI—finding hidden patterns in data that were collected for other purposes. The integration of multiple data types—imaging, circulating tumor DNA, liquid biopsies—could enable a predictive, multi-modal early warning system.”

If paired with the right algorithms, ordinary medical exams—such as an eye test, a chest X-ray or a blood panel—could double as cancer screening. In his study, Madabhushi learned through RetHemo that the retina can serve as an early marker of systemic disease.

This approach could allow clinicians to screen for blood cancers with a simple, non-invasive eye exam, opening yet another door to earlier interventions and better outcomes. "We're not simply saying, 'Well, we found the cancer,’” Madabhushi says. “There's an opportunity to prognosticate the onset of these cancers and therefore start to build interventional strategies, lifestyle changes, modifiable risk factors. That’s where I think the real big opportunities are.”

BIG BLOOD BET

All this work exists against the backdrop of GRAIL. The biotech giant’s Galleri multi-cancer early detection (MCED) blood test looks for DNA “fingerprints” shed by more than 50 cancers. The test, which is prescription-only, is aimed at older adults (50 and over) or those at elevated risk. In the U.S., the PATHFINDER 2 study has enrolled more than 35,000 participants at 30 institutions, including Morehouse School of Medicine. In the U.K., the trial has completed its final study visits with more than 140,000 participants, with results due in 2026.

Galleri isn’t FDA-approved, but it has “breakthrough device designation,” and findings from these trials will support the test’s premarket approval application. Kwong is rooting for success. “I think what GRAIL is doing is enormously important,” he says. “If they succeed, it’s a game-changer. And if they don’t, that teaches us something, too. Either way, it pushes the field forward. We need both the big companies and the academic labs to be trying everything we can.”

---

Jerry Grillo is a veteran journalist and two-time Georgia Author of the Year nominee.

---