This startup CEO says he can ‘help you lead a healthy life to 100’

Predictive AI CEO Yun Sa-jung poses after an interview with the Korea JoongAng Daily at the media outlet’s office in western Seoul on Sept. 4. [PARK SANG-MOON]

[GAME CHANGER]
 
What if an AI knew your body better than you do?
 
Predictive AI, a Korean-founded medical startup that has raised $13.8 million domestically, is betting on that future. The company is building digital twins of the human genome — virtual replicas of some 20,000 genes and three billion DNA sequences — that can predict disease, flag risky prescriptions and even step in as a patient’s proxy during consultations. The vision, says CEO and co-founder Yun Sa-jung, is simple but bold: Help people stay healthy, to age 100, by keeping every organ in peak condition.
 
“It’s not about living forever,” Yun told the Korea JoongAng Daily in a recent interview at the media outlet’s office in western Seoul. The CEO, who also serves as an adjunct professor of bioinformatics at Johns Hopkins University, added, “It’s about ensuring each organ runs smoothly for as long as possible.”
 
The inspiration came from a personal turning point. When Yun’s father was diagnosed with a rare stomach cancer, he and his identical twin brother, Yun Si-jung — now Predictive AI’s chief science officer and a research professor of genomics at Johns Hopkins School of Nursing — realized the disease could have been anticipated, and possibly prevented, through genetic testing.
 
That regret catalyzed a four-year effort to turn the human genome into a digital twin. In 2020, the brothers launched Predictiv Care in Silicon Valley with early backing from Plug and Play Ventures, later expanding to the United Arab Emirates with support from Mubadala Investment Company. More recently, they created a Korean entity, Predictive AI, for their entry into the domestic market.
 
Now the startup is racing to prove the real-world potential of its platform. By simulating the body at a genetic level, the system can recommend the safest drug combinations for patients, avoiding harmful interactions. It can act as a tireless “medical advocate,” providing doctors with precise personal health data that patients themselves may not be aware of. And through liquid biopsy — a simple blood draw — it aims to detect cancers earlier than conventional methods like colonoscopy, offering a less invasive and more frequent option for preventive care.
 
The following are excerpts from the interview, edited for length and clarity.
 
 
Q. What are your flagship products, and how do they help patients?
A. We focus on two main solutions. The first is BestMed, which helps match patients with the right medications based on their genetic profile. For instance, some medications are prodrugs that require activation by liver enzymes. But not everyone carries the same enzymes, which means the drug may fail or cause side effects in certain patients.
 

A brochure for Predictive AI’s flagship product, BestMed [PREDICTIVE AI]

 
If a patient takes a drug their body can’t metabolize, it won’t work — or worse, it could cause serious side effects. By simulating drug-gene and drug–drug interactions through a genetic digital twin, we can recommend the safest and most effective prescriptions for each individual.
 
The second is CancerTracer, which uses liquid biopsy to detect fragments of tumor DNA circulating in the blood. Traditional methods like colonoscopies or protein-based cancer markers are useful, but they often detect disease only once it has progressed. By contrast, liquid biopsy has the potential to catch cancers earlier, even at very small concentrations, and in a much more convenient way, simply from an extra tube of blood during a checkup. Together, these tools are designed to give people preventive, personalized care that extends their life span while preserving health.
 
How long does it normally take to copy a person’s genetic blueprint digitally?
From the moment we collect a cheek swab or blood sample, it takes about 60 hours to complete sequencing. A buccal swab contains saliva and epithelial cells, and extracting the DNA from that material takes roughly eight hours. Once prepared, the sample goes into the sequencing machine, which requires another 50 hours to generate the full genetic sequence.
 
But the bigger challenge isn’t the speed. It’s the cost structure. Sequencing machines from companies like Illumina or MGI use cartridges that process samples in bulk, whether it’s one or hundreds. Because of this, much of the wait time comes not from the sequencing itself but from holding samples in cold storage until enough are collected to run a full batch.
 
 
It’s interesting that you chose to combine pharmacogenomics and liquid biopsy. How did that come to be?
For us, it felt natural. Once you’ve built the digital twin, it’s like building Google — you can integrate different applications on top of it. Pharmacogenomics and liquid biopsy look at different aspects of the genome. One examines inherited germ line mutations to see how your body processes drugs; the other looks for somatic mutations — changes that occur in cells over time like those that drive cancer. Both rely on DNA, so combining them on the same platform made sense.
 
The idea is not to replace doctors, but to give them a precise advocate for the patient. A digital twin can answer questions patients themselves often can’t, like whether they carry a gene linked to breast cancer. It can also run simulations in real time: if a cardiologist prescribes one drug and a pulmonologist prescribes another, the system checks for genetic and chemical conflicts before the patient even fills the prescription. In that sense, it acts like a medical proxy — helping optimize care organ by organ, drug by drug.
 

A brochure for Predictive AI’s flagship product, Cancer Tracer [PREDICTIVE AI]

 
What are your plans for Korea and how do you expect your solutions to be incorporated into Korea’s medical system?
Korea is a very attractive market because of its national health screening program. Right now, checkups are offered in standardized packages. But medicine isn’t one size fits all. With genetic information, we can move toward personalized screenings. For example, someone with a higher genetic risk for Parkinson’s disease may need imaging from a specific angle that a standard CT scan doesn’t cover. If we know that in advance, we can tailor the checkup accordingly, saving time and money while catching diseases earlier.
 
Of course, regulations here are stricter than in the U.S. In the U.S., direct-to-consumer genetic testing is broadly permitted, whereas in Korea. access is more limited due to tighter regulations. That means most of our work has to flow through hospitals and clinics, rather than directly to consumers. Still, I believe our solutions can become a natural extension of Korea’s health system to turn routine screenings into truly preventive personalized care.
 

Predictive AI CEO Yun Sa-jung poses after an interview with the Korea JoongAng Daily at the media outlet’s office in western Seoul on Sept. 4. [PARK SANG-MOON]

You also have entities in the United States and UAE. What are your plans for overseas?
Each market is unique. In the U.S., our work has centered on seniors, who often take five or more medications at once. Polypharmacy is a huge issue, and BestMed provides a clear solution by helping physicians manage drug safety. In the UAE, we were selected for the Hub71 accelerator with support from Mubadala Group.
 
Ultimately, our long-term vision is global. We want to demonstrate a one-of-a-kind clinic that embodies what we call the ‘5P medicine’ — personalized, preventive, predictive, participatory and precision care. The idea is that anyone who walks into such a clinic can have all their vital organs fine-tuned back to normal range, lowering risk factors before they develop into disease. If we can achieve that, people everywhere will have a better chance of living healthily to 100.
 

BY LEE JAE-LIM [[email protected]]