Using aberrant methylation patterns for tumor classification and prediction of malignant progression

Brain tumors of similar malignancy grades and histology often have a vastly different clinical and biological course. Classifying tumors based on large scale analyses of DNA or RNA is a promising approach for distinguishing similar appearing tumors. Aberrant methylation of CpG islands in DNA will be particularly useful in classifying tumors, since it plays an important role in inappropriate silencing of cancer genes in sporadic human tumors, including low grade brain tumors in which genetic alterations are far less frequent. Here we propose that the aberrant methylation is intimately associated with and contributes to the malignant behavior of brain tumor cells. A prediction based on this hypothesis is that gliomas of different malignancy grades may exhibit aberrant methylation patterns that are specific to a grade. The hypothesis also suggests that aberrant methylation might allow prediction of future behavior of tumor cells, prior to recurrence. We will test our hypothesis and predictions by determining the methylation status of 1184 CpG islands in low grade and high grade gliomas, using Restriction Landmark Genome Scanning (RLGS). RLGS is a method that separates radiolabeled NotI restriction in two dimensions, in a reproducible and quantitative fashion. As a prerequisite to tumor classification and prediction, we will first identify each of the 1184 genes corresponding to the anonymous CpG islands displayed on the RLGS profile, using our arrayed CpG island library. To classify the tumors, and to allow prediction of malignant progression based on methylation patterns, we will apply statistical methods and clustering techniques that are well suited for analysis of large data sets. Predicting progression at the time of the primary tumor diagnosis is critical to the appropriate application of aggressive therapy and to avoiding unwarranted toxicity. Successful prediction strategies will present a therapeutic window for preventing malignant dedifferentiation that is potentially months to years in duration.