Two-Transcript Gene Expression Classifiers In Diverse Phenotypes
Lucas Edelman, Department of Bioengineering, University of Illinois at Urbana-Champaign, 119 Roger Adams Lab, MC-712, Box C-3, 600 S. Mathews Avenue, Urbana, IL 61801 and Nathan D. Price, Department of Chemical and Biomolecular Engineering, University of Illinois at Urbana-Champaign, 119 Roger Adams Lab, MC-712, Box C-3, 600 S. Mathews Avenue, Urbana, IL 61801.
The advent of high-throughput DNA microarray technology has rendered genome-wide transcriptional analysis a standard practice in biological research. We describe here a simple classifier that reliably discriminates between binary phenotypic states with just two gene expression measurements. Rather than summing cumulative expression across many genes, we employ an algorithm which compares the relative expression of all possible pairs of genes in a set, and selects those gene pairs which exhibit an order of expression correlated strongly with phenotype- with a relative expression ratio greater or less than 1 dependent on the class of the sample. This algorithm is invariate to data normalization and generates robust, statistically significant classifiers even with low sample sizes. Microarray comparison of the gastrointestinal malignancies GIST and LMS generated a two-transcript classifier validated in the clinic and decisive for effective therapeutic planning. We demonstrate novel, highly accurate two-transcript classifiers in diverse phenotypes, including other cancers, cardiomyopathies, viral infection, and differentiation. These simple bivariate analyses represent a powerful modality for the stratification of cellular and pathological processes, and embody a robust analytical tool for clinical diagnostics.