Rice Oligonucleotide Array (ROMA) Gene Expression Database Data Processing Methods FAQ
1. How are the images quantified?
The TIFF images are quantified using Genepix 5.1 (Axon Instruments, Union City, CA). Both the Cy3 and Cy5 image are analyzed simultaneously. Using a GAL-file (Gene Array List) the grid is overlaid on the image. Initially, print blocks of the array are identified automatically by the software and adjusted manually were needed. After block alignment the features within the blocks are identified automatically by the software. The software automatically flags spots that cannot be found in one of the channels by assigning a flag value of -50. The raw intensities of the quantified image are saved in a gpr file. Using a Perl script, additional spots are flagged that do not meet the following criteria: spots containing > 30% saturated pixels in either channel, spots with a distorted diameter in either channel and spots that could not be validated during the microarray production process. These spots are all assigned a flag value of -100. The background intensity is calculated by the Genepix software as follows: the median pixel intensity is calculated from a circular region with three times the diameter of the spot, excluding the pixels assigned to neighboring spots. Median background intensity is used to reduce the effect of spurious pixels contributing to the background.
2. What is the background correction and data normalization process?
For background correction and data normalization for each hybridization the raw intensities are loaded into the limma package of Bioconductor (www.bioconductor.org) using the read.maimages function. For the Cy5 channel the F635 Mean column of the gpr-file is used as foreground intensity and the B635 Median column is used as background intensity. For the Cy3 channel the F532 Mean and B532 Median columns are used as foreground and background intensities. Spots with a negative Flag value (Flags column in the gpr file) are assigned a weight of 0 using the wt.fun function of the limma package. Background subtraction and normalization is performed by the normalizeWithArrays function of limma. Background intensities are subtracted from the foreground intensities and negative values are set to 0. Background corrected intensities are normalized by the printtiploess method using default paramaters. Because the flagged spots are assigned a weight of 0, these spots are excluded from the normalization process (and loess curve fitting). The normalized and background subtracted intensity values are exported from limma both for the red (Cy5) and green (Cy3) channel, the flagged spots are assigned a value of "NA".
Last modified: Wednesday, 02-Nov-2005 15:59:43 EST