The Duke-NUS Breakthrough in AI-Driven Rare Disease Care

In a groundbreaking perspective published in PLOS Medicine, researchers from Duke-NUS Medical School in Singapore have proposed a transformative framework for using artificial intelligence (AI) to overhaul care for people living with rare diseases. This study emphasizes organizing AI applications around the patient journey, introducing a patient–clinician–AI triad that spans early detection to personalized therapies. With rare diseases affecting an estimated 5% to 8% of Singapore's population—or roughly 300,000 to 480,000 individuals—this work positions Singapore's higher education institutions at the forefront of global health innovation.

The paper, led by Joanne Michelle D'Souza and colleagues at Duke-NUS, highlights how AI can address the notorious 'diagnostic odyssey' that plagues rare disease patients, often lasting 5 to 7 years globally. By leveraging Singapore's advanced digital health infrastructure, this research not only promises faster diagnoses but also paves the way for more equitable care in a multi-ethnic society.

Rare Diseases: A Hidden Burden in Singapore

Rare diseases, defined as conditions affecting fewer than 1 in 2,000 people, collectively impact millions worldwide, with over 7,000 identified types. In Singapore, conservative estimates suggest 2,000 to 3,000 individuals live with chronic rare diseases, while broader genetic analyses indicate 3.4% to 8% prevalence for rare genetic disorders. Children are disproportionately affected, comprising about 700 cases, and 30% may not survive past age five without intervention.

Singapore's diverse population—Chinese, Malay, Indian, and others—complicates diagnosis due to varying genetic risks. For instance, a SingHealth Duke-NUS Institute of Precision Medicine (PRISM) pilot mined 1.28 million electronic health records (EHRs) to uncover undiagnosed cases of familial hypercholesterolemia and Fabry disease, demonstrating data analytics' power in this context. Yet, challenges persist: limited local data, specialist shortages, and high treatment costs strain the system, underscoring the urgency for AI solutions.

Duke-NUS Medical School: Leading AI Research in Singapore

Duke-NUS, a collaboration between Duke University and the National University of Singapore (NUS), exemplifies Singapore's commitment to translational research. Its Centre for Biomedical Data Science and initiatives like CARE-AI focus on ethical AI deployment, including tools for rare disease detection via EHR mining. Recent contributions include involvement in RareArena, the largest benchmark dataset for large language models (LLMs) in rare disease diagnosis, covering 72,661 cases across thousands of conditions.

The school's Ophthalmology and Visual Science program contributed to RareArena, where GPT-4o achieved 64.2% top-1 recall in confirmation tasks, outperforming other LLMs—particularly for genetically inherited diseases. This positions Duke-NUS graduates as leaders in higher ed jobs blending medicine and AI.

The Patient-Clinician-AI Triad: A New Paradigm

Central to the Duke-NUS study is the patient–clinician–AI triad, a collaborative model restructuring AI around four journey stages:

• Early Detection: AI analyzes EHRs, wearables, and public health data for surveillance, flagging anomalies like unexplained symptoms.
• Diagnosis: Multimodal AI integrates genomics, imaging, and phenotypes to shorten odysseys.
• Clinical Trials: AI matches patients to trials via natural language processing (NLP) of eligibility criteria.
• Therapies: Predictive modeling tailors treatments, forecasting responses and side effects.

"Artificial intelligence (AI) can transform rare disease care when organized around the patient journey," the authors state, advocating clinician oversight to ensure human judgment prevails.

AI-Powered Early Detection and Surveillance

AI excels in spotting rare diseases proactively. In Singapore, PRISM's EHR analysis identified hidden familial cases, a model scalable via machine learning. Globally, AI processes vast datasets from wearables to predict outbreaks or individual risks, reducing missed diagnoses from 30-50%.

For example, anomaly detection algorithms flag atypical lab results, prompting referrals. Duke-NUS's transfer learning adapts high-resource models to local data, as shown in cardiac arrest predictions (80% accuracy in Vietnam). This is vital for Singapore's aging population and multi-ethnic risks.

Transforming Diagnosis with Multimodal AI

Diagnosis is AI's stronghold. RareArena benchmarks LLMs like GPT-4o at 33-64% recall for screening/confirmation, excelling in systemic diseases. Singapore's Genomics for Kids ASEAN has cut diagnosis from 7.6 years to weeks (52% success), now building a regional registry.

Tools like face2gene use imaging/genomics for phenotyping, while NLP extracts insights from unstructured notes. Duke-NUS envisions integrated platforms prioritizing clinician validation.

Streamlining Clinical Trials Access

Only 5% of rare disease patients join trials due to matching hurdles. AI uses NLP to parse criteria against patient profiles, as in TrialGPT prototypes boosting matches 2-3x. Singapore's Rare Disease Project Registry facilitates model-human researcher pairings, enhanced by AI.

In Asia, where trials lag, this democratizes access, accelerating therapies for underserved groups.

Personalized Therapies and Predictive Modeling

AI forecasts drug responses via pharmacogenomics, optimizing therapies. For instance, NUS's CURATE.AI adjusted doses for rare cancers using small data. Duke-NUS advocates AI for monitoring adherence and adverse events, tailoring plans dynamically.

Overcoming Challenges: Data, Equity, and Ethics

Data scarcity hampers AI; federated learning and synthetic data offer solutions. Equity concerns in multi-ethnic Singapore require diverse training sets. CARE-AI at Duke-NUS develops bioethics tools for fairness.

• Regulatory gaps: Need clinician-AI governance.
• Integration: Upskilling via med schools like Duke-NUS.
• Trust: Transparent models vital.

Singapore's Ecosystem: Registries and Initiatives

The Rare Disease Fund supports 2,000+ patients. Registries like Singapore Rare Disease Project and ASEAN Genomics for Kids leverage AI for matching.Explore careers in genomic research. MOH and A*STAR drive AI-health synergy.

Future Outlook: AI's Role in Singapore's Health Future

By 2030, AI could halve diagnostic times. Duke-NUS calls for global consortia like POLARIS-GM. Singapore leads via talent from NUS, NTU, Duke-NUS.

Photo by Ortopediatri Çocuk Ortopedi Akademisi on Unsplash

Career Opportunities in AI and Rare Disease Research

Singapore universities offer booming roles in AI-medicine. Research assistant jobs at Duke-NUS blend data science and clinical work. Rate professors on Rate My Professor for insights. Check higher ed career advice for paths in precision medicine. Explore university jobs and faculty positions.