The gene coexpression analysis in RiceFREND is performed based on 24 datasets representing 815 microarray data including redundant data among the datasets. Details on each dataset can be accessed from the RiceXPro database. The data have also been deposited in NCBI’s Gene Expression Omnibus (GEO), and accessible through GEO series accession numbers, GSE21396, GSE21397, GSE36040, GSE36042, GSE36043, GSE36044, GSE39423, GSE39424, GSE39425, GSE39426, GSE39427, GSE39429, and GSE39432. All data were generated using the rice 4x44K microarray RAP-DB platform, which contains 35,760 independent probes corresponding to 27,201 annotated loci published in RAP-DB.

The microarray data were obtained by either one-color (intensity-based) or two-color (ratio-based) hybridization system. For data sets using the one-color method, the processed raw signal intensity of each probe was subjected to 75th percentile normalization and transformed to log2 scale. In addition, the median expression value across the data within each dataset were subtracted for each probe. In the case of multiple probes corresponding to the same locus, the average expression value was used. For microarray data obtained by two-color microarray system, the expression value for each gene was calculated as the log-ratio of signal intensity (log2 Cy5/Cy3).

The microarray data from all datasets were combined into one gene expression matrix data. The weighted Pearson’s correlation coefficient (PCC) was calculated to reduce sample bias that may result from using more data from one organ/tissue sample such as leaf. The mutual rank (MR) was used as an index for coexpression as described in ATTED-II.