Ó³»­´«Ã½

Researchers at Dana-Farber Cancer Institute and the Ó³»­´«Ã½ have developed an artificial intelligence-based diagnostic tool that could transform the way acute leukemia is identified and treated. The tool, called MARLIN (Methylation- and AI-guided Rapid Leukemia Subtype Inference), uses DNA methylation patterns and machine learning to classify acute leukemia with speed and accuracy. This tool has the potential to significantly improve patient care by allowing faster and more precise treatment decisions.

Acute leukemia is an aggressive blood cancer that requires accurate diagnosis to guide treatment. Current diagnostic methods, which rely on a combination of molecular and cytogenetic tests, often take days or even weeks to complete. Built as a neural network, MARLIN, however, can provide results in as little as two hours from the time of biopsy. By providing rapid and detailed insights into leukemia subtypes, MARLIN could enable clinicians to make treatment decisions sooner and with more complete information.

"Ultimately, we envision that methylation-based acute leukemia classifications will complement standard-of-care diagnostic tests to provide more comprehensive and timely information to pathologists, clinicians, and patients," said , a computational biologist at Dana-Farber, an associate member in Ó³»­´«Ã½'s Epigenomics Program, and co-senior author of the study. 

The findings were published in .

To develop MARLIN, the research team first created a large reference database of DNA methylation patterns from over 2,500 patient samples, representing a wide range of leukemia subtypes. This revealed 38 distinct methylation classes across disease subtypes in adults and children. "Our analysis confirms many established molecular categories in acute leukemia but also reveals new classes that are best seen through the lens of epigenetics," said Til Steinicke, co-first author of the study along with Salvatore Benfatto.

Using this reference dataset, the team trained MARLIN to identify methylation classes in bone marrow and blood samples from patients with acute leukemia. When paired with long-read nanopore sequencing technology, which permits DNA methylation profiling from clinical samples, MARLIN demonstrated high accuracy for disease classification in both retrospective and prospective samples. "We adapted our machine learning approach to use very small amounts of data as input, which allowed us to quickly generate classifications after just a few minutes of sequencing," said Benfatto.

In real-time tests, MARLIN correctly classified leukemia samples in under two hours from the time of biopsy receipt, a dramatic improvement over the timelines of current diagnostic workflows. "As our work continues, we hope MARLIN will greatly accelerate treatment planning, reduce medical complications, and ease patient worries in the future," said Evan Chen, a collaborator on the study and medical oncologist at Dana-Farber.

The researchers also found that MARLIN could resolve diagnostic "blind spots" missed by conventional methods. For example, it successfully identified cryptic genetic events such as rearrangements involving the DUX4 gene that are associated with favorable clinical outcomes. Additionally, MARLIN revealed novel predictive signatures, such as HOX-activated subgroups, which could inform future treatment strategies.

"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 Gabriel Griffin, a cancer researcher, Ó³»­´«Ã½ associate member, Dana-Farber pathologist, and co-senior author of the study.

The researchers now aim to develop MARLIN into a tool that can be provided to patients in the clinic to accelerate disease classification, inform treatment selection in real-time, and generate a resource for the cancer research community to study the role of DNA methylation in leukemia.

Adapted from issued by Dana-Farber Cancer Institute.