Esteller maps methylation classifiers
- Blood Advances published a multinational study led by Annapurna Saksena and colleagues, with Manel Esteller among the authors, describing a DNA methylation classifier for hematolymphoid neoplasms built from 1,156 tumor samples. - The team identified 44 reproducible methylation classes across more than 1,400 cases overall, and its machine-learning classifier reached 97% diagnostic concordance when it made high-confidence calls. - The paper says classifier results changed some real-world diagnoses, extending methylation profiling beyond brain tumors into blood cancers. (ashpublications.org)
DNA methylation is a chemical tagging system on DNA, and a paper published April 23 in *Blood Advances* says those tags can help sort blood cancers into diagnostically useful groups. (ashpublications.org) The study, led by Annapurna Saksena and colleagues with Manel Esteller among the co-authors, assembled what it called the largest cross-platform methylome cohort yet for hematolymphoid neoplasms: 1,156 profiled samples. (ashpublications.org) Across more than 1,400 hematolymphoid neoplasms in the full analysis, the researchers defined 44 reproducible methylation classes that tracked closely with World Health Organization 5th edition and International Consensus Classification disease entities. (ashpublications.org) (omicsdi.org) A methylation classifier is a pattern-matching tool: it compares a new tumor’s DNA-tagging profile with reference profiles and returns the closest match. In this study, the machine-learning model showed 97% diagnostic concordance in cases where it produced a high-confidence result. (ashpublications.org) (omicsdi.org) The authors said several methylation signatures split established tumor types into smaller subclasses with statistically distinct patient outcomes. They also reported that higher tumor purity made high-confidence classifier matches more likely. (omicsdi.org) That matters because diagnosing lymphomas, leukemias and related cancers can be difficult when tissue is limited or the pathology is ambiguous. The paper describes the method as a proof of concept for a future clinical tool rather than a finished stand-alone test. (ashpublications.org) The study also says classifier output changed diagnoses in specific clinical cases, a sign that the assay can add information beyond standard pathology. The associated dataset record says an external online portal for the hematolymphoid tumor methylation classifier is planned. (omicsdi.org) Methylation profiling is already used in some other cancers, especially central nervous system tumors, and the authors frame this work as an attempt to bring the same logic to blood and lymphoid malignancies. Esteller’s group has also been building related methylation reference resources in hematologic cancer cell lines. (ashpublications.org) (carrerasresearch.org) The immediate result is not a new drug or guideline. It is a bigger reference map for reading the DNA “fingerprints” of blood cancers, one diagnosis at a time. (ashpublications.org)
