Cherbas Lab
Research:
Microarray analysis is a powerful technique for analyzing gene expression changes. The DGRC manufactures transcriptome microarrays for the Drosophila community and we have found them very effective. Each DGRC microarray is a glass slide with DNA spots (synthetic oligonucleotides) corresponding to every Drosophila gene we know (see the photo below). To use it one isolates RNAs, labels them (as cDNAs) adding fluorescent tags, then carries out what amounts to a dot blot experiment. Each slide is hybridized simultaneously to cDNAs from 2 RNA samples, each tagged with a different fluorescent dye.
The challenging part is analyzing all the data. This has been an active area of investigation and there are a number of popular approaches. We have developed a new technique -- ArrayLOD -- that seems to us to have striking advantages.
ArrayLOD takes advantage of the strong and reproducible dependence of technical variation (scatter from all sources other than actual differences in RNA titer) on the intensity of the fluorescence signal of a microarray spot. The relationship between technical variation and signal intensity is determined experimentally through "self-hybridizations" to yield a fitted standard curve. The significance of experimental measurements is assessed by reference to this standard curve.
Data from biological replicates on separate slides are combined by a maximum likelihood approach. Significance level for each gene is reported through its False Discovery Rate.
The results of an ArrayLOD analysis differ in several ways from those yielded by most commonly used procedures. Because it provides an accurate estimate of technical error, ArrayLOD yields fewer false positives from low-intensity spots and makes it possible to detect very small effects (20-30% differences in RNA titer) from strongly expressed genes.
ArrayLOD can separate technical and biological variation, allowing detection of genes whose responses (expression ratios) vary among biological samples. Such genes are reported and flagged, instead of being automatically discarded.
We have used ArrayLOD to detect differentially expressed genes in a study of the ecdysone response of Kc cells in Drosophila melanogaster. We obtained excellent results from 3 biological replicate slides, and even a single slide gave reliable (though sparse) data.