Rice Microarray Data Integration and Analysis

As of March 2012, microarray data from 105 rice microarray experiments (1,867 hybridizations) were collected from NCBI GEO (Barrett et al.[2009 (link)]), EBI ArrayExpress (Parkinson et al.[2007 (link)]) and PLEXdb (Dash et al.[2012 (link)]). The raw data was downloaded and experiments without raw data were discarded. For one-channel array (Affymetrix), MAS 5.0 method provided by the R package, affy, for the Affymetrix rice array was used to conduct background correction, normalization, probe specific background correction, probe summarization and convert probe level data to expression values (Affymetrix[2012 ]). The trimmed mean target intensity of each array was arbitrarily set to 500. The data were then log₂ transformed. For two-channel arrays (Agilent 22K and 44K, BGI/Yale, NSF 20K and 45K), R package marray in Bioconductor was used to do the normalization with within-array Lowess and between-array MAD scale normalization methods (Cleveland[1979 (link)]; Wang et al.[2002 (link)]). The color-swap hybridizations were manually corrected to make them comparable among other samples. In case of Agilent 44K array data used for meta-profiling analyses, we converted the median signal intensities of Cy3 to log₂ median intensities and then normalized the log₂ intensities using the quantile normalization method (Bolstad et al.[2003 (link)]). The sequences of probes were extracted from each platform website and then mapped onto RGAP V6, RAP V3 and KOME cDNAs using NCBI Blast with 100% identity over 100% coverage (Altschul et al.[1990 (link)]). Regarding Affymetrixprobeset which have 11 probe pairs, the probeset with at least half perfect-match (PM) probes matched onto cDNA sequence was considered as mapped.
The ROAD database was constructed with PHP (Hypertext Preprocessor) and MySQL, run on a Windows 2003 server. The http address ishttp://www.ricearray.org. Heatmap and classic line plots were generated by the PHP library JpGraph (http://jpgraph.net/).