Hope springs eternal in medicine. Picture a time, perhaps within the next 10 years, when the idea of a “digital health twin” becomes real. A virtual model that shows how a person’s body might respond to different treatments. When we not only know what’s wrong with someone but also what might go wrong and how to prevent it.

The Human Phenotype Project (HPP), a large-scale, international study designed to track individuals for up to 25 years to map the complete journey from health to disease, could provide the dataset needed to build advanced AI models that shift healthcare from reactive treatment to proactive prevention and toward a personalized health ecosystem.

IMPACT IN THE MIDDLE EAST

This project is notable because the UAE plays a key role. The Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) is the scientific co-leader of the HPP, providing advanced AI frameworks for processing massive, heterogeneous datasets.
Researchers at the university are developing foundation models, such as GluFormer, to predict diabetes risk, demonstrating how a technology-driven health system can improve healthcare outcomes through data and AI.

“Over the next 10 years, the HPP will drive a paradigm shift in the region from reactive treatment to proactive health management,” says Eran Segal, Dean of the Biological and Life Sciences Division and Professor of Computational Biology at MBZUAI, who heads the HPP.

“In the Middle East, where diseases such as obesity, diabetes, and cardiovascular disease are highly prevalent, phenotypic clustering of dense multimodal data is particularly powerful,” adds Segal.

Because these conditions develop gradually and are influenced by a complex interplay of biology, environment, and behavior, Segal adds, they cannot be fully understood from static snapshots, such as a single blood test. The HPP is well-suited to tackle these conditions because “it has created some of the largest and most detailed datasets on heart and metabolic health.”

COMPLEMENTING THE GENOME PROGRAM

So, how does the HPP differ from genome-focused projects like the Emirati Genome Program?

“To understand the HPP, it is helpful to first define a phenotype: while your genome is the static blueprint of DNA you’re born with, your phenotype consists of the observable traits—clinical, biological, and behavioral—that manifest as your genes interact with your environment and lifestyle over time,” says Segal.

More than 25 years ago, scientists reached a major milestone: the first draft of the human genome. The sequence of chemical “letters” in our DNA has led to important medical advances.

In the past few years, the Emirati Genome Program and Abu Dhabi Biobank have made progress in disease prevention, early detection, and precision medicine.

The HPP supports these genome-focused efforts. While the Emirati Genome Program provides foundational insights into inherited risks, the HPP provides the real-world context of how those risks manifest, says Segal.

“By treating genetics as one of 30 different data layers, the HPP helps explain the dynamic interplay between a person’s DNA and their daily life, enabling a more complete version of precision medicine.”

The HPP gathers detailed data that many genome projects miss. This includes data from wearables and imaging, such as CGM, sleep studies, eye scans, and bone density scans, as well as detailed molecular profiles of the gut and mouth microbiomes.

By tracking these data layers over time, Segal says the project trains AI models to distinguish between a person’s actual age and their biological age.

“This allows for the creation of a personalized digital twin or a computational model that can simulate medical interventions and forecast health trajectories before a disease even manifests.”

CHALLENGES IN SCALING DEEP PHENOTYPING

Scaling deep phenotyping can be challenging because of the strategic trade-offs between study depth and participant scale.

Deep profiling takes more time and resources. For example, collecting over 30 types of data—from diet to molecular “omics”—requires significant engineering and advanced infrastructure. Combining different types of data, like images, sensors, and tests, is also technically difficult.

Participants’ commitment to a 25-year study is important.  “While it doesn’t require frequent clinic visits, it helps to think of the HPP as a long-term health portfolio: you only check in with a clinic visit every two years, but the continuous data collected in between builds a 25-year wealth of insights that allows AI to forecast your future health,” says Segal

Managing large, diverse datasets also requires strict rules to keep data secure and ensure benefits are shared fairly.

LEADING THE NEXT GENERATION OF BIOBANKING

Expanding the project in the UAE helps to ensure it includes an ethnically diverse cohort. “This diversity allows the HPP to establish refined, personalized medical norms that apply to global populations, and moves beyond the ‘one-size-fits-all’ diagnostic approach,” says Segal.

For decades, Segal says modern medicine has been designed using a template that doesn’t fit the whole world. “Today, medicine has a ‘data divide’. Most of what we know about genetics comes from a very small group of people,” he says.

That’s why the UAE is an ideal location for biobanks. It has people from the Middle East, Africa, and South Asia—groups often missing from global research data.

“Through the success of the Emirati Genome Program, which is the world’s largest genomic database, the UAE has already proven why it’s uniquely positioned to lead the next generation of biobanking,” says Segal.

Diversity, he adds, becomes especially powerful when paired with longitudinal phenotyping, as it allows researchers to observe how risk, resilience, and aging unfold differently across populations over time, rather than simply how they differ at a single point.

But diversity alone isn’t enough. “It needs to be combined with careful recruitment, strong community involvement, ethical oversight, and fair access to research benefits to ensure responsible and widely applicable science,” he says.

PRECISION MEDICINE NEEDS PHENOTYPING

Genomics gives an important blueprint of inherited risk, but true precision medicine needs detailed phenotyping.

Dr. Niyas Khalid, Specialist in Internal Medicine at Burjeel Hospital in Abu Dhabi, says precision medicine without full phenotyping “risks becoming algorithmically impressive but biologically incomplete. Genomics identifies biological susceptibility, and phenotype defines clinical expression.”

Dr. Khalid explains, “Managing rare metabolic disorders such as acute hepatic porphyria, we frequently encounter patients with pathogenic variants who remain asymptomatic due to variable penetrance. Conversely, we see individuals with compelling biochemical and clinical phenotypes in whom intronic variants or variants of uncertain significance become clinically important, because the phenotype is compelling.”

Ignoring phenotype when examining genotype introduces uncertainty, adds  Dr. Khalid.  “Precision emerges when phenotype anchors interpretation and genotype justifies it. In rare disease medicine, this becomes even more nuanced; some phenotypes may not be genetically confirmed. In such cases, clinical decision-making must be guided by phenotype, not by relying on genetic confirmation.”

Traditional medicine often depends on single clinical visits, which are snapshots of health. Deep phenotyping, however, tracks over 30 data layers, including CGM, the microbiome, and advanced imaging to capture real-time biological changes.

High-resolution phenotyping can reveal risks that standard tests miss, says Segal. “For instance, HPP research showed that 40% of individuals with normal fasting glucose levels were reclassified as prediabetic when tracked continuously using CGM. This depth of insight is simply not achievable with static genetic or clinical measures alone.”

Precision medicine also needs to act on data before the disease develops. “The HPP’s AI foundation model, GluFormer, recently published in Nature, uses deep glycemic data from just two weeks of CGM monitoring — over 10 million measurements from more than 10,000 participants — to outperform the current standard of care, such as HbA1c blood tests,” says Segal.

In a landmark 12-year study, GluFormer demonstrated strong predictive power, identifying 69.2% of cardiovascular deaths and 65.8% of new diabetes cases in its highest-risk group, with none in the lowest-risk group. The traditional HbA1c test, he adds, did not show meaningful differences for these long-term risks.

Experts agree the HPP could change clinical practice in the region over the next decade, shifting healthcare from waiting for symptoms to using predictive biological models.
“Rather than diagnosing only when glucose exceeds a defined number or cardiac function declines beyond a cutoff, clinicians could identify dynamic risk signatures earlier,” says Dr. Khalid.

He adds that phenotype-driven modeling may soon be essential for providing high-quality, personalized care. “If implemented strategically, the region could transition from high disease burden to high-resolution biological surveillance.”