Researchers have developed an artificial intelligence model that uses DNA methylation patterns to identify and classify blood cancers.

These cancers include numerous subtypes that can be difficult to distinguish using conventional diagnostic approaches, yet accurate classification is essential for determining prognosis and treatment.

This work was led by Pedro Orsini, M.D., of ALMA Genomics.

Francisco Marchi, Ph.D., the corresponding author, presented the research at the Association for Molecular Pathology 2026 Annual Meeting & Expo.

The conference explores the emerging role of AI in molecular diagnostics, including identifying difficult-to-classify cancers, improving the accuracy of AI-generated reports and helping laboratories use genetic testing more efficiently.

The researchers (from ALMA Genomics, the University of Pennsylvania, the Massachusetts Institute of Technology and several other academic and clinical institutions) developed ALMA 3, an AI foundation model trained using a large collection of (more than 110,000) patient samples.

In independent testing, the diagnostic model detected the presence of hematolymphoid tumours with 99.5% sensitivity and achieved a 90.9% positive predictive value when assigning specific cancer subtypes.

The study included 81 subtypes of blood and lymphatic cancers, including several that are particularly difficult to classify using conventional diagnostic methods.

The findings demonstrate the potential for AI-assisted molecular classification to complement conventional pathology, although additional clinical validation is under way.

Source: Association for Molecular Pathology