Artikel
DNA methylation array-based molecular profiling for brain tumor classification
Suche in Medline nach
Autoren
Veröffentlicht: | 25. August 2015 |
---|
Gliederung
Text
Question/methods: Molecular technologies which can complement standard pathology testing have the potential to greatly enhance diagnostic precision. Analysis of DNA methylation, acting as a 'fingerprint' of cellular origin, is one such promising technology for tumor classification. We have established a brain tumor DNA methylation profile reference bank consisting of more than 2,000 cases using the Illumina HumanMethylation450 (450k) array, with over 70 brain tumor entities or subgroups represented. The array platform works from frozen or paraffin-embedded material, with minimal DNA input. A random forest algorithm has been established to compare diagnostic cases with the reference cohort resulting in an entity prediction with an associated probability score. Genome-wide copy number profiles and target gene methylation data (e.g. MGMT) generated from the array provide additional diagnostic information.
Results: We have analysed more than 1000 diagnostic samples from Heidelberg University Hospital and external institutions with this classification scheme. A considerable number of cases displayed a discrepancy between histological and molecular diagnoses. The rate of discrepancies varies between entities and interestingly also between age groups. For example the rate of discrepancies is relatively low for glioblastomas above the age of 30 years (n=183, discrepancy in 3%) but considerably higher in patients below 30 years (n=48, discrepancy in 24%). Careful re-examination of these cases often resulted in refinement of the original diagnosis, in several cases resulting in a change of suggested patient care. Furthermore, we have collected evidence of further subgroups of certain entities and associations with recurrent copy number changes. These subgroups await further characterization.
Conclusions: We present the establishment and feasibility of diagnostic implementation of a DNA methylation based brain tumor classification scheme. By the analysis of over 1000 prospective diagnostic cases we can demonstrate that the utility of this method varies from entity to entity or between age groups within specific entities. In our experience, for many cases DNA methylation profiling provides an assuring confirmation of histological diagnosis. For cases with discrepant results additional diagnostic workup and reconsideration of the histological diagnosis is highly advocated. We further demonstrate that for many unclear histological cases additional diagnostic information can be derived from methylation profiles.