Power and accuracy of QTL detection: simulation studies of one-QTL models.

摘要

Non-parametric selective resampling procedures were used to investigate the effect of several factors on the power of quantitive trait loci (QTL) detection, the bias and confidence intervals of their position, effect and heritability. The factors studied were population size, QTL position, heritability of the QTL and marker coverage, i.e., marker density and their regular versus random spacing. Confidence intervals obtained using either Normal approximation, the bias corrected and accelerated (BCa)method and empirical bootstrap with 1000 selected resamples were compared. The BCa intervals were found to be very close to classic confidence intervals (CI) assuming normal distribution for sample sizes above 200, and to be slightly closer to empiricalCI for small population sizes. The precision of the QTL position was found to be mostly affected by population size and heritability, and less by marker spacing, except in the case of sparse maps with irregular marker spacing. Bias in QTL position estimates can be high for small population sizes when QTL are located near the end of a chromosome, and, unexpectedly, selective bootstrap does not decrease this bias very much.