ABSTRACT
Most advanced imaging techniques yield enormous quantities of data, sometimes at great expense, and often representing complex spatiotemporal associations across experimental groups. Oncology is arguably the hallmark application of microarray technology. Indeed, many of the earliest microarray applications involved cancer cell lines or biopsy material. These experiments highlight the great promise of this technology, while underscoring the necessity for meticulous experimental design and analysis. The publication of Golub’s microarray experiments on leukemic cells was a seminal event; these experiments demonstrated accurate computer-based recapitulation of the well-known classification of leukemia into T-cell and B-cell subtypes, based solely on microarrayexpression signatures. In addition to diagnostic applications, many microarray experiments have been designed to determine whether differences in gene expression reflect differences in prognosis, rather than alternative genetic pathways to a similar tumor histology. Chen et al. used decision trees to derive gene-expression signatures that predict survival in non–small cell lung cancer.
