Hitchhiking mapping: A population-based fine-mapping strategy for adaptive mutations inDrosophilamelanogaster

Abstract

lnRV test statistic was incorrect for locus MS3L-11 (last line of Table 2). The conclusion that two regions of significantly reduced variability are located on the X chromosome and the third chromosome remains unchanged. In the right-hand portion of Fig. 1A, the curve shown was also incorrect. The corrected table and figure appear below.

Controlling the False-Positive Rate in Multilocus Genome Scans for Selection

“…First, in practice, one obtains P-values for the CLRT by simulating a null distribution under the standard neutral model (Harr et al 2002;Kim and Stephan 2002;Bauer-DuMont and Aquadro 2005;Beisswanger et al 2006;Pool et al 2006), which is problematic when the demographic assumptions of that model are violated ( Jensen et al 2005). Second, the CLRT (and the subsequent GOF) are not applied to randomly chosen loci in practice, but to outlier loci identified by a genome scan (e.g., Harr et al 2002;Bauer-DuMont and Aquadro 2005;Beisswanger et al 2006;Pool et al 2006). Such ascertainment procedures choose loci with very unusual underlying genealogies, resulting in a pattern of spatial variability that may mimic that of a recent selective sweep, such as an excess of highfrequency, derived mutations surrounding a region of reduced diversity (Figure 4).…”

“…Of particular importance is how such loci are chosen from the tails in the first place. In practice, the choice is not made using the results of the CLRT/GOF (due to computational impracticability), but rather on the basis of a summary of the data (e.g., Harr et al 2002;Glinka et al 2003;Bauer-DuMont and Aquadro 2005;Beisswanger et al 2006;Pool et al 2006). Teshima et al (2006) have recently found that choosing outlier loci on the basis of low levels of diversity is more powerful than choosing on the basis of summaries of the site-frequency spectrum.…”

“…Where whole-genome variation data are unavailable, investigators will sample levels of variability from multiple regions of the genome using markers that are both relatively rapid and relatively inexpensive to type, such as microsatellites (e.g., Harr et al 2002;Kauer et al 2003;Bauer-DuMont and Aquadro 2005), or short fragments of nucleotide sequence to identify singlenucleotide polymorphisms (SNPs) (e.g., Glinka et al 2003;Tenaillon et al 2004;Ometto et al 2005;Wright et al 2005). From such studies, a subset of these regions may then be selected for additional sequencing, and the parameters of a model of recent positive, directional selection acting on new mutations will be estimated from the data.…”

“…For example, genome-scan data consisting of short reads of DNA sequence may be summarized by the number of mutations in each fragment, with invariant fragments being used to identify regions for further sequencing (e.g., Glinka et al 2003;Schlenke and Begun 2004;Beisswanger et al 2006). Similarly, a region may be identified because variability and/or allele frequencies of microsatellite markers are extremely skewed in some regions of the genome relative to the data set as a whole (e.g., Harr et al 2002;Bauer-DuMont and Aquadro 2005;Pool et al 2006). The rationale for the follow-up experiment is that the statistics used to identify outlier regions (e.g., Tajima 1989; Voight et al 2006) are not formal tests for selection, as they do not specifically reject a neutral model in favor of a model including selection.…”

“…In Jensen et al (2005), the calculation of the GOF test was applied to simulated data assuming that loci are random draws from a population model. In practice, however, both the method of Kim and Stephan (2002) and the GOF tests are often applied to loci that are preselected by an investigator because some feature of the region is an outlier in a multilocus genome scan (e.g., Harr et al 2002;Bauer-DuMont and Aquadro 2005;Beisswanger et al 2006;Pool et al 2006).…”