New Tool Could Revolutionize Acute Leukemia Diagnosis

Current diagnostic methods, which rely on molecular and cytogenetic testing, can take several days or even weeks to deliver results. This delay can complicate care for patients who need immediate intervention. Now, researchers have developed a new approach to deliver accurate classifications in just a few hours, potentially transforming leukemia care.

Researchers at Dana-Farber Cancer Institute (Boston, MA, USA) have created a diagnostic tool called MARLIN (Methylation- and AI-guided Rapid Leukemia Subtype Inference). The system uses DNA methylation profiling and machine learning to classify leukemia subtypes with unprecedented speed. Built from a reference database of over 2,500 patient samples representing 38 methylation classes, MARLIN leverages epigenetic signatures to capture disease complexity across adult and pediatric cases.

The tool was trained using a neural network capable of analyzing bone marrow and blood samples with high precision. Combined with long-read nanopore sequencing, MARLIN generates accurate classifications using minimal data in just minutes of sequencing. This integration enables clinicians to obtain diagnostic insights within two hours of a biopsy, significantly faster than current workflows.

In tests, MARLIN correctly classified retrospective and prospective leukemia samples with high accuracy. The findings, published in Nature Genetics, demonstrated that MARLIN could resolve diagnostic blind spots missed by conventional techniques. It successfully identified cryptic genetic events, such as DUX4 rearrangements, and uncovered novel predictive signatures, including HOX-activated subgroups, that could inform treatment strategies.

In addition to diagnosis, MARLIN holds promise for advancing treatment planning and reducing complications by accelerating decisions. Its ability to uncover new epigenetic subtypes suggests broader applications in predictive biomarker development and drug response monitoring. The team now aims to integrate MARLIN into clinical workflows, where it could serve as a real-time decision support tool and as a resource for studying leukemia biology.

“We believe that our framework paves the way for future developments in epigenetic classification of acute leukemia, machine learning-assisted diagnostics, and methylation-based predictive biomarkers of drug response,” said Dr. Gabriel Griffin, co-senior author of the study.

Related Links:
Dana-Farber Cancer Institute