A focused clinical programme designed to build predictive knowledge from the population it is intended to serve.
Predict Health Lab proposes a 24-month clinical programme focused initially on Type 2 diabetes, designed to build and validate an Explainable AI (XAI) model capable of identifying disease-progression patterns earlier and with greater resolution than approaches based on routine clinical data alone. To do this, the programme goes beyond conventional clinical information (glucose, HbA1c, lipid profile): it incorporates advanced biological biomarkers — gut microbiome, cellular and blood oxidative stress, telomere length and epigenetic markers — that capture early physiopathological dimensions not systematically represented in routine health records, together with longitudinal follow-up of each participant.
The UAE already has one of the most advanced health-data and population-screening infrastructures in the world. The additional step Predict Health Lab proposes is to equip that infrastructure with biological data of the highest possible resolution, drawing on the full extent of medical and research knowledge available today — not only the data that routine clinical practice alone can offer.
Prospective · Observational · Longitudinal
Proposed cohortUP TO1,000PARTICIPANTS
Phase I24MONTHS
Initial clinical focusType 2DIABETES
Illustrative image · AI-generated
The objective is not simply to create another health dataset.
IT IS TO LEARN HOW RISK DEVELOPS OVER TIME — AND WHETHER THAT KNOWLEDGE CAN SUPPORT EARLIER CLINICAL ACTION.
01 / The foundational clinical pool
UP TO 1,000 PARTICIPANTS. DEEPLY CHARACTERISED. FOLLOWED OVER TIME.
01
Routine clinical data
Glucose regulation, metabolic indicators, cardiovascular risk factors and other routinely available clinical variables.
02
Advanced biological data
Selected molecular and biological biomarkers designed to reveal mechanisms that routine clinical information alone may not capture.
03
Longitudinal data
Repeated measurements and clinical evolution over time.
Illustrative image · AI-generated
The value lies not simply in the number of participants, but in the depth and longitudinal quality of the information generated.
A deeply characterised study cohort, not a statistically representative sample of the entire UAE population.
02 / From biology to prediction
SEE DEEPER. LEARN OVER TIME. EXPLAIN THE RISK.
CLINICAL DATA + ADVANCED BIOMARKERS + LONGITUDINAL FOLLOW-UP
EXPLAINABLE AI
PATTERNS & TRAJECTORIES
CLINICAL VALIDATION
Explainability is fundamental. The aim is not to produce an opaque risk score, but to develop predictive intelligence whose contributing factors can be examined, interpreted and evaluated in a clinical context.
Explainable Artificial Intelligence is intended as clinical decision support. Physicians retain responsibility for clinical decisions within the healthcare system.
03 / From a focused cohort to population-level potential
DEEP PHENOTYPING DOES NOT NEED TO MEAN MASS TESTING.
The foundational pool creates two complementary layers of information: advanced biological measurements and routine clinical data. By studying them together over time, the programme can investigate which patterns in routine clinical records correspond to the higher-risk biological trajectories identified within the deeply characterised cohort.
If clinically validated, these relationships can be translated into analytical risk criteria using variables already present within the healthcare system. Those criteria could then be applied, under UAE governance, to larger health datasets to identify populations that may benefit from earlier evaluation or prioritisation.
DEEP PHENOTYPING
LEARN THE TRAJECTORIES
IDENTIFY CANDIDATE PATTERNS
CLINICALLY VALIDATE
TRANSLATE THE LEARNING
ANALYSE EXISTING HEALTH DATA
PRIORITISE RISK
EARLIER CLINICAL EVALUATION
ONE FOCUSED COHORT. A POTENTIAL FOUNDATION FOR POPULATION-LEVEL PREDICTIVE HEALTH.
Risk prioritisation is not diagnosis. Application to existing data would require appropriate validation and UAE health-system governance.
04 / Validation before scale
PROVE FIRST. SCALE SECOND.
The programme is deliberately structured as a validation phase. It does not assume national implementation before evidence exists.
Did the model identify meaningful predictive patterns?
Were those patterns clinically valid?
Can they be applied responsibly to existing health data?
Does the approach justify broader implementation?
Phase IBUILD → LEARN → VALIDATE → MEASUREEvidence and evaluation
UAE DECISION
Potential next phaseINTEGRATE → PRIORITISE → EXPANDSubject to successful validation
Predict Health Lab is seeking an institutional dialogue with the UAE to explore the proposed Predictive Health Programme and the framework required for its clinical validation.
