It’s human nature to categorize the world around us. From the earliest taxonomic classification defining us as Homo sapiens, we’ve grouped dogs by breed (pugs, bulldogs, retrievers), wines by grape (merlots, chardonnays, pinot noirs), and of course, cancers by origin (gliomas, melanomas, sarcomas). You can order a DNA test for your dog, or ask the sommelier about your wine, but how can we ensure the right tumor classification?
Currently, the World Health Organization (WHO) primarily classifies central nervous system (CNS) tumors based on histological methods. This is subject to the pathologist’s interpretation, which inevitably introduces variability and error. Alternatively, in-depth molecular profiling is needed. Given that the consequences for misclassifying a tumor are much higher than ordering the wrong wine, researchers out of the German Cancer Research Center and University Hospital Heidelberg (Heidelberg, Germany) developed an unbiased CNS tumor classification system based on tumor DNA methylation profiles. Here’s how they did it:
- First, they established a reference cohort using the 450k BeadChip Array to acquire genome-wide methylation data for ~2,800 tumor samples. The samples were derived from nearly all current WHO classifications, as well as a variety of tumor microenvironments
- Unsupervised iterative clustering gives rise to 82 classes of CNS tumors based on unique methylation patterns
- They used the reference cohort to develop a diagnostic tool to sort and classify 1,104 patient samples; 977 of the samples successfully matched to a methylation class
- 86% of matches are in agreement between the histopathological approach and the methylation approach
- In 171 of these, a specific molecular subgroup could be identified, which is not possible using histology only
- For 139 samples, the identified DNA methylation class is not in agreement with histopathological diagnosis. These samples were further evaluated with in-depth molecular diagnostics
- In 129 of 139 cases, the histopathological diagnosis is incorrect, and was resolved in favor of DNA methylation class
- In 71% of the revised diagnoses there was also a change in the WHO grading (a measure of tumor malignancy)
This work holds promise as a more accurate system for the diagnosis of CNS tumors, which could in turn improve treatment plans and patient outcomes.
To get the details on how they developed this methylation-based classification system, click over to the online classifier tool and check out Nature, March 2018