The positions of these QTLs were estimated from the physical positions of either two flanking markers (the genomic positions of the distal ends of the two markers) or a co-segregated single marker (between the start and end positions on the genome). By screening for redundancy of traits at the same physical positions, representative QTLs were selected.

The data were normalized by data processing, including PRIM (processing implementation for microarray) filtering, and ‘sector median’ normalization, performed at each spotting sector on the glass slide to remove uneven sectors. All the expression data are shown as values of the expression ratio.