Molecular voting for glioma classification reflecting heterogeneity in the continuum of cancer progression

摘要: Gliomas, the most common brain tumors, are generally categorized into two lineages (astrocytic and oligodendrocytic) further classified as low-grade (astrocytoma oligodendroglioma), mid-grade (anaplastic astrocytoma anaplastic high-grade (glioblastoma multiforme) based on morphological features. A strict classification scheme has limitations because a specific glioma can be at any stage of continuum cancer progression may contain mixed Thus, more comprehensive molecular signatures reflect biological nature tumors accurately. In this study, we used microarray technology to profile gene expression 49 human applied k-nearest neighbor algorithm for classification. We first trained set with 19 typical cases selected genes that provide lowest cross-validation error k=5. then 30 remaining cases, including several do not belong gliomas such atypical meningioma. The results showed only does correctly classify gliomas, but detailed voting also subtle information regarding similarities neighboring classes. For meningioma, was equally split among four classes, indicating difficulty in placement meningioma classes gliomas. actual results, which typically decide winning class label algorithms, useful method gaining deeper insight tumor development.