Methylation classifier for sarcomas

What it measures. Every cell type carries a methylation pattern set during development, and tumours keep the pattern of the cell they came from plus changes of their own. A methylation array reads hundreds of thousands of sites from a small amount of DNA extracted from routine formalin-fixed tissue. A random forest classifier trained on more than a thousand reference sarcomas (Nature Communications 2021, from the Heidelberg and DKFZ groups) returns the most likely class with a calibrated score, and copy number changes come free from the same array.

Who should have it. Sarcomas are rare, have more than seventy subtypes, and are misclassified more often than common cancers even by expert pathologists. The classifier is most useful for undifferentiated or small round cell tumours, for tumours in unusual sites, for children and young adults, and whenever the reference pathologist's diagnosis and the molecular findings disagree. It is run at reference centres and through the free online classifier at molecularneuropathology.org, where a laboratory uploads its own array data.

What changes. A confident class can change the diagnosis, and with it the chemotherapy protocol (Ewing sarcoma versus other round cell tumours, for example), the expected behaviour, and eligibility for trials or targeted drugs. A low score is reported as unclassifiable rather than forced, so the method does not replace histology and fusion testing; it sits alongside them. The classifier is an academic research tool without regulatory clearance; the array costs a few hundred pounds per sample and turnaround is one to two weeks.