Simple sequence repeat marker associated with a natural leaf defoliation trait in tetraploid cotton.
The detection of quantitative trait loci (QTL) requires large sample sizes to attain reasonable power (Soller et al. 1976). For reduction of the number of individuals needed to be genotyped in markerQTL linkage experiments, a procedure termed "selective genotyping" has been proposed for experimental species (Darvasi and Soller 1992; Lander and Botstein 1989; Lebowitz et al. 1987) and has been adapted for humans as well (Risch and Zhang 1995). With selective genotyping, only individuals from the high and low phenotypic extremes are genotyped. It has been shown that the number of individuals genotyped to attain a given power can be decreased significantly, at the expense of a moderate increase in the number of individuals phenotyped (Darvasi and Soller 1992). The major limitation of this approach is that if the experiment is aimed at analyzing a number of traits, then by selecting the extremes of each trait one would select most of the population and thus no reduction in genotyping can be obtained. Selective genotyping is thus most appropriate for the cases where only one trait is being analyzed. This conclusion is valid when selective genotyping is applied to QTL detection. However, after a QTL is detected, its map location will be estimated. In this case, additional markers at chromosomal regions of interest will be used to provide better estimation of QTL map location, since for QTL detection alone, relatively wide marker spacing is adequate (Darvasi and Soller 1994). These additional markers will be at a specific chromosomal region, and thus they will ordinarily concern a single specific QTL and a single specific trait only. Consequently, with respect to these markers, selective genotyping can be applied, even if the initial experimental population was used to map QTL affecting several traits. The proportion of the population selected for genotyping with the additional markers will be determined by the effect of selective genotyping on QTL mapping accuracy. This aspect of selective genotyping is the object of the present study. The model used to investigate the effect of selective genotyping on QTL mapping accuracy consists of a 100-cM chromosome with a single QTL located at its center. A backcross population, originating from crosses between two inbred lines with alternative alleles for all markers and for the QTL, is assumed. The quantitative trait is taken to have a normal distribution with equal variance, cr 2, for all QTL genotypes; and standardized gene effects of 0 and d for the QTL genotypes Qq and QQ respectively. Darvasi and colleagues (1993) have shown that QTL mapping accuracy when using an infinite number of markers is the same as that achieved by interval mapping using moderate marker spacing. Consequently, simulation results obtained on the assumption of an infinite number of markers apply for the actual experimental case where markers spaced at fairly wide intervals are used. On this basis and for simplicity of calculations only, an infinite number of markers, represented in the simulations by a marker every 0.1 cM, is assumed. Monte Carlo simulations were carried out according to the above model, and a maximum likelihood estimate (MLE) was obtained for QTL map location, as detailed in Darvasi and colleagues (1993). A 95% empirical confidence interval (CI) was obtained from 1000 replicated simulations for each parameter combination. Heron, CI is referred to as the length of the 95% confidence interval. The equivalence of various experimental designs (BC, F2, halfsibs) when using selective genotyping has also been presented (Darvasi and Soller 1992). Consequently, the present model serves as a close approximation to a wide range of actual experimental conditions. Previous studies have shown that CI is mainly a function of sample size, N, and gene effect, d (Darvasi et al. 1993). In order to explore the independent effect of N and d, six different parameter combinations of N and d were chosen: N = 500 with d = 2.0, 0.7 and 0.5; and N = 2000 with d = 1.0, 0.35 and 0.25. The values of d were chosen to provide similar CI with the two different sample sizes. Confidence intervals were estimated selecting a total proportion, p, of the population (p/2 at each phenotypic extreme). Figure 1 presents CI as a function of the proportion selected, p, for the six parameter combinations. As expected, CI increases when smaller proportions of the population are selected. It can be seen that population size and gene effect, independently, do not substantially affect the way CI changes as a function of p. That is, the influence of proportion selected on CI will be similar for any particular combination of d and N that determines the same CIs when selecting the entire population. Most importantly, it can be seen that in all cases, selecting more than 40-50% of the population does not reduce the CI.