|Application:||Gene expression microarray analysis|
|Number of samples:||10|
|Release date:||Oct 1 2007|
|Last update date:||Mar 16 2012|
|Dataset link||Gene expression non-additivity in immature ears of a heterotic F1 maize hybrid|
We directly contrasted both B73 and H99 inbred lines vs. their F1 heterotic hybrid transcriptomes in immature ear. Our experimental design consisted of 10 cDNA microarray hybridizations, 5 for each combination of hybrid vs. inbred genotypes, involving 20 separate labeling reactions. Labeling dyes were swapped in two of the five replicates for each combination. In each hybridization, control channel was assigned to the F1 hybrid, and differences in transcriptional levels between B73 and H99 were inferred using the hybrid as a common reference sample in an indirect experimental design. To take in account variability in transcript population among individuals, total RNA coming from different isolations, each collected on multiple individuals, were mixed before poly(A+) RNA purification. All hybridizations on microarray slides were then performed using cDNA independently labeled from the poly(A+) RNA purification product for each genotype. Base-two logarithms of expression ratios were subjected to one-class response significance analysis in SAM v. 2.20 software [Tusher et al. 2001]. For each EST the estimates of additive parameter "a" and dominance parameter "d" (middle-parent heterosis) were obtained as a = (L2 – L1)/2 d = (L1 + L2)/2 (where L1 and L2 are mean base-two logarithms of expression ratios of F1 vs. B73 and F1 vs. H99, respectively). Positive values of "a" indicate expression values bigger in B73 than in H99. The dominance/additivity ratio (d/|a|) was also calculated [Falconer 1989]. Evaluation of statistical significance of parameters was done by calculating standard errors of the estimates "a" and "d" as standard errors of linear functions of the means. Significance testing was done correspondingly, using an F-test for linear contrasts. P-values for the families of tests corresponding to each parameter were subjected to global error analyses using a method based on fitting mixture distribution [Allison et al. 2002], allowing to estimate the false discovery rates (FDR) and false negative rates (FNR). Confidence intervals for d/|a| ratios were obtained by Fieller’s method [Piepho and Emrich 2005], allowing to classify the genes into different dominance type classes.