Researchers at the LKS Faculty of Medicine of the University of Hong Kong (HKUMed) have developed a groundbreaking artificial intelligence tool that promises to revolutionize cardiovascular disease (CVD) risk assessment, potentially identifying individuals at high risk of serious cardiac events up to 15 years before the onset of symptoms. This innovative system, named CardiOmicScore, leverages the power of multiomics and deep learning, analyzing a comprehensive suite of biological signals from a single blood test to predict the future likelihood of six major cardiovascular conditions.
The implications of this development are profound, offering a paradigm shift from reactive treatment to proactive prevention in the global fight against cardiovascular disease, which remains the leading cause of death worldwide. In 2022 alone, CVDs accounted for an estimated 19.8 million fatalities, underscoring the urgent need for more effective early detection strategies.
The Science Behind CardiOmicScore: A Multiomic Approach
Cardiovascular disease risk is traditionally assessed using a combination of clinical factors such as age, blood pressure, cholesterol levels, smoking history, and family predisposition. While valuable, these methods often fail to capture the subtle, early biological changes occurring within the body long before any outward symptoms manifest. This can lead to a critical delay in diagnosis and intervention, by which time the window for optimal preventive measures may have significantly narrowed.
Genetic risk scores, such as polygenic risk scores, offer another avenue for assessing predisposition. These scores aggregate the influence of numerous genetic variants to provide a measure of inherited susceptibility. However, an individual’s genetic makeup is largely fixed from birth, limiting their ability to reflect dynamic changes influenced by lifestyle, diet, environmental exposures, aging, and other modifiable factors.
CardiOmicScore addresses this critical gap by providing a more current and dynamic snapshot of an individual’s biological health. The system employs advanced deep learning algorithms to integrate multiple layers of biological information – a process known as multiomics. This approach synergistically combines data from genomics (the study of genes), proteomics (the study of proteins, which perform most of the body’s functions), and metabolomics (the study of small molecules called metabolites, which are byproducts of metabolic processes and can indicate the body’s current state of health and disease).
The HKUMed team analyzed extensive population data from the UK Biobank, a large-scale biomedical database. Their model meticulously examined 2,920 circulating proteins and 168 metabolites present in blood samples. These molecules collectively provide a detailed picture of a person’s current biological state, potentially revealing early shifts in immune activity, metabolic function, and vascular health that precede the development of overt symptoms.
Professor Zhang Qingpeng, Associate Professor in the Department of Pharmacology and Pharmacy at HKUMed and a leading figure in this research, explained the synergy of this approach: "Genes determine where we start – they define our baseline health risk. However, proteins and metabolites reflect our current physical health. Our AI tool is designed to decode these complex molecular signals, enabling doctors and patients to identify risks much earlier, which can potentially change the trajectory of disease through timely lifestyle modifications and early prevention."
Predicting Six Major Cardiovascular Diseases
The CardiOmicScore system has demonstrated remarkable efficacy in predicting the future risk of six major cardiovascular diseases: coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease, and venous thromboembolism. These conditions collectively represent a significant burden of morbidity and mortality globally.
Coronary artery disease, characterized by the narrowing or blockage of the arteries supplying blood to the heart, is a primary driver of heart attacks. Stroke, caused by a disruption of blood supply to the brain, can lead to permanent disability or death. Heart failure occurs when the heart cannot pump blood effectively to meet the body’s needs. Atrial fibrillation is an irregular heartbeat that significantly increases the risk of stroke. Peripheral artery disease affects circulation in the limbs, often causing pain and impaired mobility. Venous thromboembolism involves dangerous blood clots forming in veins, which can dislodge and travel to the lungs.
In individuals identified as being at elevated risk by the CardiOmicScore, the model was able to detect warning signals as far back as 15 years before clinical symptoms would typically emerge. This predictive power significantly surpasses that of conventional polygenic risk scores. The researchers found that the accuracy of CardiOmicScore further improved when traditional clinical information, such as age and gender, was incorporated into the model.
The findings, published in the prestigious scientific journal Nature Communications, highlight the potential of multiomic data, interpreted through advanced AI, to offer unprecedented insights into an individual’s future health trajectory.
A Timeline of Discovery and Development
The journey leading to the development of CardiOmicScore represents a significant investment in scientific research and technological innovation. While the exact timeline of the HKUMed research project is not fully detailed in the initial release, the process typically involves several key stages:
- Conceptualization and Data Acquisition (Early Stages): The initial phase would have involved conceptualizing the multiomic approach to cardiovascular risk prediction. Access to large-scale, well-characterized population datasets like the UK Biobank is crucial for training robust AI models. This stage likely spanned several years, involving ethical approvals, data standardization, and the collection of comprehensive biological samples and clinical data.
- Model Development and Training (Mid-Stages): The core of the research involved developing and training the deep learning algorithms. This is an iterative process where researchers feed the AI model vast amounts of multiomic and clinical data, along with information on whether individuals later developed specific CVDs. The AI learns to identify complex patterns and correlations that are invisible to human analysis. This phase could take several years, involving experimentation with different algorithmic architectures and feature selections.
- Validation and Refinement (Later Stages): Once a preliminary model was developed, rigorous validation was essential. This involved testing the model on independent datasets to ensure its predictive accuracy and reliability. Refinements would have been made based on validation results to optimize performance. The reported ability to predict up to 15 years in advance suggests extensive testing across different time horizons.
- Publication and Dissemination (Present): The publication of findings in Nature Communications signifies the culmination of years of rigorous research and peer review. This marks the formal introduction of CardiOmicScore to the scientific community and the broader public, paving the way for future clinical translation.
The research team, led by Professor Zhang Qingpeng, also includes members from the HKU Musketeers Foundation Institute of Data Science (IDS), with Luo Yan identified as the first author. This collaborative effort underscores the interdisciplinary nature of modern biomedical research.
Broader Implications: Shifting Towards Proactive Health Management
The development of CardiOmicScore signifies a critical step in the evolution of precision medicine. While traditional genetic testing offers a static view of inherited risk, multiomic tools like CardiOmicScore provide a dynamic and evolving assessment of an individual’s health status. This dynamic capability is paramount, as our biological landscape is constantly influenced by external and internal factors.
The potential impact on public health is immense. By shifting the focus from treating established diseases to predicting and preventing them, healthcare systems can potentially reduce the incidence of costly and debilitating cardiovascular events. For individuals, early and accurate risk prediction empowers them to make informed lifestyle choices, engage in closer medical monitoring, and potentially benefit from early therapeutic interventions.
The ability to predict risk up to 15 years in advance is particularly transformative. This extended timeframe provides a substantial window for implementing preventive strategies. Imagine an individual receiving a CardiOmicScore indicating a high risk of stroke in their early 50s. Armed with this knowledge, they could proactively work with their healthcare providers to manage blood pressure, adopt a heart-healthy diet, increase physical activity, and undergo regular check-ups, significantly reducing their actual risk of experiencing a stroke.
Professor Zhang further articulated the long-term vision: "We aim to leverage technology to identify and prevent diseases before they develop. By shifting health management from reactive treatment to proactive prediction and intervention, we aim to create a lasting impact for both public health and individual patient care."
Challenges and Future Directions
While the potential of CardiOmicScore is undeniable, several challenges lie ahead before it can be widely adopted in clinical practice. These include:
- Clinical Validation and Regulatory Approval: Extensive clinical trials will be necessary to validate the predictive accuracy and clinical utility of CardiOmicScore across diverse populations. Regulatory bodies will need to review and approve the technology for use in healthcare settings.
- Cost-Effectiveness and Accessibility: The cost of multiomic profiling and AI analysis needs to be evaluated to ensure it is accessible to a broad patient population. Efforts will be required to make this advanced diagnostic tool economically viable for healthcare systems.
- Integration into Clinical Workflows: Seamless integration of CardiOmicScore into existing clinical workflows will be essential for its practical implementation. This will involve training healthcare professionals on how to interpret and act upon the results.
- Ethical Considerations: As with any advanced predictive technology, ethical considerations surrounding data privacy, genetic discrimination, and the psychological impact of risk prediction will need to be carefully addressed.
Despite these challenges, the development of CardiOmicScore represents a beacon of hope in the ongoing battle against cardiovascular disease. It exemplifies the power of artificial intelligence and multiomics to unlock new frontiers in medical diagnostics, paving the way for a future where diseases are detected and managed long before they can inflict their most devastating consequences. The research from HKUMed offers a compelling glimpse into a future of healthcare that is more predictive, personalized, and ultimately, more effective in safeguarding human health.
