We found a total of thirty-five main-effect genomic regions and many epistatic interactions controlling genistein, daidzein, glycitein and total isoflavone accumulation in seeds.
Indeed, in W2, the main effects of markers explained more than 60% of the genomic variance for pairs of environments having sample phenotypic correlations greater than 0.33 (Table 4); on the other hand, in the two environments showing a low correlation (0iFN + 5iBN), the main effect explained only about 10% of the genomic variance.
Finally, in W1, the pair of environments with the largest sample phenotypic correlation (0iBN + 2iBN) had estimates of variance components such that the main effect explained about 80% of the total genomic variance.
On the other hand, the pairs of environments showing correlations larger than 0.3 (2iBN + 5iBNZ and 2iBN + 5iFN) gave estimates of variance components where the main effect explained between 50 and 70% of the genomic variance.
Furthermore, the estimates of variance components from the M×E model indicated that the proportion of genomic variance explained by the main effect of markers is directly related to the (sample empirical) phenotypic correlation between environments.
For instance, in W1, the analysis of pairs of environments exhibiting sample phenotypic correlations smaller than 0.3 (0iBN + 5iBNZ, 0iBN + 5iFN, and 5iBNZ + 5iFN) yielded estimates of variance components in the M×E model where the main effect explained less than 50% of the total genomic variance, computed as the sum of the main effect plus interaction variance estimates (see Table 3).
In the interaction models, the total genomic variance can be partitioned into a main effect and an interaction component.
Combined main effect and intervention interaction analyses raise novel hypotheses concerning the MRPS30 genomic region and the effects of hormonal and dietary exposures on postmenopausal breast cancer risk.
A main genomic mutation parameter is estimated that describes the average number of synonymous changes genome wide, and then departures from this main effect at the gene level are explored in the framework of a log linear model.
Our most general model has the form, log R = α + T + C + G + T × G, or equivalently, log E = log L + α + T + C + G + T × G, where T denotes the main effect of SSR type, C denotes the effect of chromosome, G is the main effect of genomic location category, and T × G allows for type by region interaction, that is, differential type effects by genomic category.
This view of the data provided a clear focus on two genomic regions, the FGFR2 region of chromosome 10 q, which has a very strong main effect along with suggestive evidence for interaction, and the MRPS30 region of chromosome 5 p, which shows evidence of a comparatively smaller main effect and suggestive evidence for interaction.
