SIGNIFICANCE THRESHOLDS FOR QTL INTERVAL MAPPING TESTS

The problem of mapping quantitative trait loci (QTL) using genetic mar ker information is of great interest to the mapping community. There a re many statistical methods available for detecting and/or locating QT L, all of which depend on assumptions about the distribution of the qu antitative trait values. The distribution of the trait values is affec ted by sample size, genetic marker density, missing data patterns, env ironmental noise, etc., all of which affect the distribution of the te st statistic used to detect/locate QTL. Failure of the test statistic distribution to follow a standard statistical distribution is the subj ect of current research. In order to declare a significant QTL effect it is necessary first to understand the behaviour of the test statisti c under the null hypothesis so that a critical value may be employed. In this paper we discuss the choices available for obtaining critical values (threshold values) used in locating QTL via interval mapping pr ocedures. We investigate threshold values obtained by different means (analytical approximations and empirical) for the same level of signif icance (type I error rate) under a normality assumption (null hypothes is of no QTL). In addition, we explore the effect of deviations from n ormality of the trait values on the threshold value by comparing analy tical approximations and empirical threshold values for simulated back cross and F-2 experiments, along with an actual experimental F-2 data set.