No abstract

Several statistical methods which employ multiple marker data are currently available for the analysis of quantitative trait loci (QTL) in experimental populations. Although comparable estimates of QTL location and effects have been obtained by these methods, using simulated and real data sets, their accuracy and reliability have not been extensively investigated. The present study specifically examines the merit of using F2 and doubled haploid populations for locating QTL and estimating their effects. Factors which may affect accuracy and reliability of QTL mapping, such as the number and position of the markers available, the accuracy of the marker locations and the size of the experimental population used, are considered. These aspects are evaluated for QTL of differing heritabilities and locations along the chromosome.A population of 300 F2 individuals and 150 doubled haploid lines gave estimates of QTL position and effect which were comparable, albeit extremely unreliable. Even for a QTL of high heritability (10~), the confidence interval was 35 cM. There was little increase in reliability to be obtained from using 300, rather than 200, F2 individuals and 100 doubled haploid lines gave similar results to 150. QTL estimates were not significantly improved either by using the expected, rather than the observed, marker positions or by using a dense map of markers rather than a sparse map. A QTL which was asymmetrically located in the linkage group resulted in inaccurate estimates of QTL position which were seriously biassed at low heritability of the QTL. In a population of 300 F2 individuals the bias increased from 4 cM to 20 cM, for a QTL with 10~ and 2~o heritability respectively.

SUMMARYCollections of S. graminicola were made from local pearl millet crops in West Africa and India. Asexual inoculum was derived from the collections in Polythene tunnels and used to infect pearl millet cultivars from both continents. Analysis of deviance of the incidence of disease was performed on the logit scale using GLIM. The host × pathogen interactions were interpreted by use of the Finlay‐Wilkinson regression technique widely used in the study of genotype × environment interactions. While the West African collections were consistently more pathogenic than the Indian ones, West African hosts were potentially more susceptible to Indian than to West African pathogen isolates; conversely some Indian hosts were more vulnerable to West African than to Indian isolates. There was a fairly stable relationship between incidence of disease and relative pathogenicity of West African isolates but the Indian isolates were more variable in their susceptibility to changing background levels of disease. These results could influence resistance breeding strategies in pearl millet improvement programmes. The methods of analysis could be valuable for application to other complex and ill‐defined pathosystems.

No abstract

No abstract