The authors have declared that no competing interests exist.

Low-frequency SNPs found in four families or fewer account for most of the largest GWAS effects (). Lack of power likely accounts for both the failure to detect small effect GWAS SNPs at low frequency and the greater proportion of intermediate-frequency GWAS SNPs relative to the null distribution (see ). Lack of power does not help explain the over-representation of large-effect SNPs at low frequency, however. Causal variants at low and high frequencies are more likely matched by random SNPs. A causal variant present in one or 25 of the 26 families has just 26 possible incidence patterns, whereas a causal variant present in 13 families has over 10 million possible incidence patterns. Our dataset of 1.6 million SNPs is too small to tag all causal variants, and we are far less likely to tag intermediate-frequency than low- or high-frequency variants. We observe large-effect SNPs at low frequency but not at high frequency, however. One explanation is linkage: linked variants with effects in the same direction will more often be combined into a single “synthetic” effect if they are present at low frequency. Low-frequency SNPs with very large effects also have low RMIP values (), which supports this explanation: rare recombinant individuals allow separation of linked synthetic loci, but are sampled only intermittently. Because all GWAS SNPS with effects over 0.3 standard deviations in this study are found in a single family, we hypothesize that they result from linked QTL. These large effects explain a small proportion of the total phenotypic variation because their frequencies are low.

La persona con dolore cronico ha bisogno di una serie di interventi e di figure...

Stay green trait was influenced significantly by applied nitrogen treatments, genotypes (lines and crosses) and their interaction. Similar results were obtained by many authors when maize plants were cultivated under a range of soil N fertilizers. Gungula . (2005) found significant differences between N rates (0 to 120 kg N ha-1) and the tested varieties for leaf senescence percentage which an indicator of the effect of soil N on greenness. The reduction in N availability encouraged leaf senescence as shown in and . Similar results were achieved by different authors; Racjan and Tollenaar (1999) found that leaf longevity was enhanced by an increase in soil N supply. In addition, reduced N availability accelerated post flowering leaf senescence than at high N and maize inbred lines showed differences in their magnitude of response (D`Andrea ., 2006). Gungula . (2005) declared that highest percentage of leaf senescence at the lowest N-level (30 kg N ha-1), while the lowest leaf senescent percentage was recorded at 120 kg N ha-1. In their results, Borrell . (2001) established that roots of the stay green sorghum maintain greater capacity to extract N from the soil compared with the non-stay green hybrids during kernel filling. They assessed it as a consequence of the balance between N-demand by the kernel and N-supply during the kernel filling. The inbred lines showed leaf greenness reduction more than their crosses (31.0 and 24.04%, respectively) when grown at N deficiency conditions. The inbreds 4, 5, 6, 7, 9, 13 and 15 and the crosses Pioneer 3062, 9x10, 4x1, 13x16 and 13x15 were characterized by their highest greening percentage at physiological maturity and kernel filling phase at reduced N soil. The previous result can be explained based on Boràas . (2003) suggestion that delay in senescing for the previous group of genotypes during kernel filling is linked to the quantity of light received by the leaves and N availability via remobilization to actively growing kernels of maize. In addition, such genotypes maintain their green leaves longer than others and in turn represent differences in photosynthetic capacity (Gungula ., 2005).

The authors have declared that no competing interests exist.

Competing Interests: The authors have declared that no competing interests exist.

The utility of GWAS studies is contingent on their ability to predict phenotypes. In this study we show that simple additive models containing several hundred SNPs explain over 50% of the phenotypic variation in a set of 4892 RILs for most of the 13 maize morphological traits (). SNP number in these models could probably be reduced considerably without sacrificing predictive ability by removing SNPs in high linkage disequilibrium with each other . Additive model predictions are least accurate for the ear traits (cob length (CL), cob diameter (CD), ear row number (ERN)), and the flowering trait anthesis-silking interval (ASI). To investigate the nature of this apparent non-additivity, we focus on models without a family term (-top) that rely solely on GWAS SNPs to explain phenotypic differences within and between families. Most traits show ~10% greater predictive ability in the parents than in the RILs, but for cob length this difference is dramatic (~30%). We observe the opposite situation for cob diameter and ear row number: predictive ability is higher in the RILs than in the parents. Here we interpret these observations in terms of interaction effects. For cob length, additive effects detected in the RILs accurately predict parental phenotypes, so we infer that interaction effects are equally likely to enhance or mask a given QTL (their mean effect is close to zero) and they must be common enough to account for a ~20% drop in predictive ability in the RILs. For cob diameter and ear row number, additive effects detected in the RILs do not predict parental phenotypes, so we infer that parental phenotypes are caused by more complex interaction effects that are seldom recapitulated in the RILs and have little influence on additive effect sizes.