AI model for pancreatic cancer developed by Academia Sinica and NTU Hospital

Researchers at Academia Sinica and NTU Hospital have created PanMETAI, an AI-powered metabolomics platform that detects pancreatic cancer early by analyzing molecular signals with NMR technology, promising improved diagnosis and patient outcomes.

Researchers at Academia Sinica and National Taiwan University (NTU) Hospital have jointly developed PanMETAI, an advanced artificial intelligence-driven metabolomics platform designed to detect pancreatic cancer at its earliest stages. The system employs a standardized nuclear magnetic resonance (NMR) analysis to identify molecular signals associated with the presence of cancer.

Challenges in Early Diagnosis of Pancreatic Cancer

Pancreatic cancer is commonly diagnosed in advanced stages due to the lack of early clinical symptoms, which contributes to a five-year survival rate of only 13%. PanMETAI aims to improve early detection and patient outcomes by providing precise screening capabilities that can identify the disease before symptoms manifest.

Functionality of PanMETAI

The platform analyzes up to 260,000 molecular signals per individual through a standardized NMR approach. Unlike conventional diagnostic methods that rely on single biomarkers, PanMETAI constructs a comprehensive metabolic profile, capturing molecular changes from pre-cancerous conditions to early lesions. This approach addresses critical gaps in current risk assessment strategies for pancreatic cancer.

Validation and Performance

The AI model underwent extensive validation, including independent blind testing in Taiwan and additional evaluation with a European cohort in Lithuania. Key findings include:

• Taiwan cohort: Area under the curve (AUC) of 0.99, sensitivity of 93%, specificity of 94%
• Lithuania cohort: AUC of 0.93

These results indicate that PanMETAI demonstrates strong diagnostic performance across multiple populations.

Integration of Expertise

The development of PanMETAI integrates over two decades of clinical experience from NTU Hospital with Academia Sinica's extensive research in metabolomics, computational modeling, and basic science. The interdisciplinary collaboration enhances the reliability of the platform and its potential applicability in clinical settings.

Future Applications

The research team expects PanMETAI to serve as an effective screening tool for individuals at high risk of pancreatic cancer. The AI system may also evolve into a broader "Multi-Cancer Early Prediction Platform," potentially expanding its role within precision medicine and early cancer detection.

Key Contributors

• Dan-Ni Wu – Postdoctoral fellow, Genomics Research Center, Academia Sinica (first author)
• Yu-Ting Chang – Professor, NTU Hospital (corresponding author)
• Chao-Ping Hsu – Distinguished research fellow, Institute of Chemistry, Academia Sinica (corresponding author)
• Chun-Mei Hu – Assistant research fellow, Genomics Research Center, Academia Sinica (corresponding author)

Sources:

• AI model for pancreatic cancer developed by Academia Sinica and NTU Hospital
• Taiwan researchers develop AI-driven pancreatic cancer detection tool
• AI-driven metabolomics platform for early pancreatic cancer detection
• Academia Sinica and NTU Hospital develop AI model for pancreatic cancer