Fig 1.
Location of the study sites (A. alba, sites 1 to 9 and A. cephalonica, site 10) and genetic structure of A. alba populations revealed by STRUCTURE for K = 2 (top) and K = 4 (bottom). The map was created in ArcMap v.9.3 (ESRI. Redlands, CA). The European basemap is copyrighted by EUROSTATS (EuroGeographics for the administrative boundaries) and is available at: http://ec.europa.eu/eurostat/web/gisco/geodata/reference-data/administrative-units-statistical-units. The black area shows the distribution range of A. alba (according to EUFORGEN 2009, http://www.euforgen.org). Study sites IDs are described in Table 1.
Table 1.
A. alba and A. cephalonica study sites and sampling design.
Table 2.
Power of the Bayesian approaches (HBM, SBM and BAYESCAN) to detect outliers for selection from simulated data.
False-discovery rates (FDR) and false non-discovery rates (FNR) were empirically assessed under different scenarios of divergent and/or homogenizing selection, by varying the proportion of selected loci, and selection strength.
Fig 2.
Hierarchical (multi-site) outlier detection.
Result of HBM (hierarchical Bayesian approach) on the complete A. alba dataset. θElev(m) are locus-specific effects on genetic differentiation among populations belonging to different elevations. On the left, the estimated values of θElev(m) with their 95% posterior credible intervals. The markers are sorted by decreasing values of θElev(m) and the dotted lines represent the inter-quantile limits [Q1-1.5(Q3-Q1); Q3+1.5(Q3-Q1)]. On the right, the distribution of θElev(m) and the fitted normal distribution. The arrow indicates the two loci detected below the neutral background in the complete dataset under a 1% probability threshold.
Fig 3.
Within-site outlier detection.
Results of FDIST and SBM within each A. alba and A. cephalonica study sites. It is noteworthy that some outliers were detected in two study sites.
Table 3.
Consistent outliers detected twice above the neutral background (by two different approaches).
The first column describes the SNP number, the second column the SNP ID. The third column describes the study site in which the outliers were detected and the method used: FDIST (within-site coalescent method) and SBM under a 1% threshold (within-site Bayesian method). Study sites IDs are described in Table 1. The complete list of outliers detected, all methods confounded, is provided in S1 Table.
Fig 4.
Observed genotypic frequencies in the different sites at SNP 255 (genotypes CC, CT and TT). SNP 255 was detected by two within-site outlier detection methods (FDIST and SBM) in site 9 (purple line), but in no other site. In addition, significant GEAs were detected in site 9 (S2 Table).