Power and false-positive rate in QTL detection with near-isogenic line libraries

Heredity (2011) 106, 576–584 & 2011 Macmillan Publishers Limited All rights reserved 0018-067X/11 ORIGINAL ARTICLE www.nature.com/hdy Power and false-positive rate in QTL detection with near-isogenic line libraries KC Falke1 and M Frisch2 State Plant Breeding Institute, University of Hohenheim, Stuttgart, Germany and 2Institute of Agronomy and Plant Breeding II, Justus Liebig University, Giessen, Germany 1 Libraries of near-isogenic lines (NILs) were used for quantitative trait locus (QTL) detection in model species and economically important crops. The experimental design and genetic architecture of the considered traits determine the statistical properties of QTL detection. The objectives of our simulation study were to (i) investigate the population sizes required to develop NIL libraries in barley and maize, (ii) compare NIL libraries with nonoverlapping and overlapping donor segments and (iii) study the number of QTLs and the size of their effects with respect to the power and the false-positive rate of QTL detection. In barley, the development of NIL libraries with target segment lengths of 10 cM and marker distances of 5 cM was possible using a BC3S2 backcrossing scheme and population sizes of 140. In maize, population sizes larger than 200 were required. Selection for the recipient parent genome at markers flanking the target segments with distances between 5 and 10 cM was required for an efficient control of the false-positive rate. NIL libraries with nonoverlapping donor chromosome segments had a greater power of QTL detection and a smaller false-positive rate than libraries with overlapping segments. Major genes explaining 30% of the genotypic difference between the donor and recipient were successfully detected even with low heritabilities of 0.5, whereas for minor genes explaining 5 !or 10%, high heritabilities of 0.8 or 0.9 were required. The presented results can assist geneticists and breeders in the efficient development of NIL libraries for QTL detection. Heredity (2011) 106, 576–584; doi:10.1038/hdy.2010.87; published online 4 August 2010 Keywords: NIL library; near-isogenic lines; QTL detection Introduction programs. However, no studies are available that investigate the effects of the experimental design used for the development of NIL libraries or the genetic architecture of the trait under consideration on the power and rate of false positives in QTL detection. A comparison of the phenotypes of NILs with overlapping donor chromosome segments was suggested for fine mapping of QTLs (Kearsey, 2002). However, the statistical properties of QTL detection in NIL libraries with overlapping segments were not yet studied and compared with those of QTL detection with nonoverlapping segments. The objectives of our simulation study were to (i) investigate the population size required in backcrossing programs to develop NIL libraries in barley and maize, depending on the desired length of donor segments and the density of the marker map, (ii) compare the power and false-positive rate of QTL detection in NIL libraries with those of nonoverlapping and overlapping donor chromosome segments and (iii) study the power and false-positive rate depending on the number of QTLs and the size of QTL effects. A near-isogenic line (NIL) library is a set of homozygous lines that carry marker-defined chromosome segments in a common genetic background (Eshed and Zamir, 1994). These segments cover the entire genome of a donor line, and were introgressed into the genetic background of a recipient line by marker-assisted backcrossing. NIL libraries were suggested to detect quantitative trait loci (QTLs) in tomato (Lycopersicum esculentum, Eshed and Zamir, 1995) and were subsequently developed in Arabidopsis (Keurentjes et al., 2007; Törjek et al., 2008) and in a wide range of crops such as rice (Oryza sativa L., Lin et al., 1998), barley (Hordeum vulgare L., Matus et al., 2003; Schmalenbach et al., 2008), wheat (Triticum aestivum L., Liu et al., 2006), maize (Zea mays L., Ribaut and Ragot, 2007; Szalma et al., 2007) and rye (Secale cereale L., Falke et al., 2009b). The experimental designs that were used in these studies for the development of NIL libraries were based on ad hoc approaches. A first investigation on the experimental design for developing NIL libraries in rye was carried out by Falke et al. (2009a), who analyzed the effect of various selection strategies on the recovery of the recipient genome and the number of marker data points required for the marker-assisted backcrossing Materials and methods Correspondence: Professor M Frisch, Institute of Agronomy and Plant Breeding II, Justus Liebig University, Giessen 35392, Germany. E-mail: Received 13 November 2009; revised 11 May 2010; accepted 14 May 2010; published online 4 August 2010 Genetic map We investigated models of barley and maize genomes. For barley, we considered seven chromosomes of 140 cM length (nc ¼ 7, lc ¼ 1.4) and for maize 10 chromosomes of 160 cM length (nc ¼ 10, lc ¼ 1.6). Barley and maize were QTL detection with NIL libraries KC Falke and M Frisch 577 chosen because of their economical importance and their character as a model species for crops. Barley has a rather short genome and maize has a longer one. This allows investigating the effect of genome length on the development of NIL libraries. The genome was covered with equally spaced markers. Map distances ranging from 2.5 to 20 cM (d ¼ 0.025, 0.05, 0.1 and 0.2) were investigated. All markers were polymorphic between the donor and recipient. Linkage maps with evenly distributed markers were assumed, because they greatly enhance the efficiency of markerassisted backcrossing (Prigge et al., 2009) and can be constructed for model species and for most crops of economical importance. To model recombination along the chromosomes, no interference in crossover formation (Stam, 1979) was assumed and recombination frequencies were related to the corresponding map distances with Haldane’s (1919) mapping function. Haldane’s mapping function is based on a simplified model of meiosis and assumes no interference in crossover formation. Nevertheless, the model resulted in simulations that are close to reality (Prigge et al., 2008). Genomic composition of the NIL libraries The target regions to which the QTL effects should be mapped were of 10 or 20 cM length (l ¼ 0.1, 0.2). For NIL libraries with nonoverlapping segments (s ¼ 1), each NIL carried one target region. For overlapping segments, 2nc NILs carried one target region and (lc/l1) nc NILs carried two adjacent target regions. The genomic composition of the NIL libraries is illustrated for d ¼ 0.2, 0.05 and s ¼ 1, 2 using the example of one barley chromosome in Figure 1. The number of NILs per library ranged from n ¼ 49 for l ¼ 0.2 and s ¼ 1 in barley to n ¼ 170 for l ¼ 0.1 and s ¼ 2 in maize. Backcrossing scheme The development of an NIL library started with crossing a homozygous donor lin (...truncated)