Mathematical modeling of noise and discovery of genetic expression classes in gliomas.

The microarray array experimental system generates noisy data that require validation by other experimental methods for measuring gene expression. Here we present an algebraic modeling of noise that extracts expression measurements true to a high degree of confidence. This work profiles the expression of 19 200 cDNAs in 35 human gliomas; the experiments are designed to generate four replicate spots/gene with switching of probes. The validity of the extracted measurements is confirmed by: (1) cluster analysis that generates a molecular classification differentiating glioblastoma from lower-grade tumors and radiation necrosis; (2) By what other investigators have reported in gliomas using paradigms for assaying molecular expression other than gene profiling; and (3) Real-time RT-PCR. The results yield a genetic analysis of gliomas and identify classes of genetic expression that link novel genes to the biology of gliomas.