Contemporary effective population size (N e ) can be estimated using linkage disequilibrium (LD) observed across pairs of loci presumed to be selectively neutral and unlinked. This method has been commonly applied to data sets containing 10-100 loci to inform conservation and study population demography. Performance of these N e estimates could be improved by incorporating data from thousands of loci. However, these thousands of loci exist on a limited number of chromosomes, ensuring that some fraction will be physically linked. Linked loci have elevated LD due to limited recombination, which if not accounted for can cause N e estimates to be downwardly biased. Here, we present results from coalescent and forward simulations designed to evaluate the bias of LD-based N e estimates (N e ). Contrary to common perceptions, increasing the number of loci does not increase the magnitude of linkage. Although we show it is possible to identify some pairs of loci that produce unusually large r 2 values, simply removing large r 2 values is not a reliable way to eliminate bias. Fortunately, the magnitude of bias inN e is strongly and negatively correlated with the process of recombination, including the number of chromosomes and their length, and this relationship provides a general way to adjust for bias. Additionally, we show that with thousands of loci, precision ofN e is much lower than expected based on the assumption that each pair of loci provides completely independent information. Heredity (Wright, 1931), which determines the rate of evolutionary change due to genetic drift and informs the equilibrium level of genetic variation and the effectiveness of selection. N e is often much lower than census size (Frankham, 1995), demonstrating that simply counting individuals is insufficient to predict rates of evolutionary change. In addition to the number of mating individuals, N e is affected by sex ratio, variation in reproductive success, age structure, migration and other demographic factors. It is an extremely relevant metric in conservation biology, with low N e leading to inbreeding and reduced genetic diversity (Ellstrand and Elam, 1993). See Charlesworth (2009) for a primer on N e , and Wang (2005) for a review of estimation methods.Populations with smaller N e undergo more genetic drift than larger populations. This genetic drift randomly generates associations between alleles at different loci, known as linkage (or gametic) disequilibrium (LD) at a rate inversely proportional to N e . As a result, measures of LD between independently-segregating loci can be used to provide an estimate of N e (Sved, 1971;Hill, 1981;Waples, 1991). Over the past decades, many studies have leveraged data sets consisting of a few dozen loci for genetic estimates of N e (Luikart et al., 2010). While these studies continue to be useful, especially for long-running

Recent advances in population genomics have made it possible to detect previously unidentified structure, obtain more accurate estimates of demographic parameters, and explore adaptive divergence, potentially revolutionizing the way genetic data are used to manage wild populations. Here, we identified 10 944 single-nucleotide polymorphisms using restriction-site-associated DNA (RAD) sequencing to explore population structure, demography, and adaptive divergence in five populations of Chinook salmon (Oncorhynchus tshawytscha) from western Alaska. Patterns of population structure were similar to those of past studies, but our ability to assign individuals back to their region of origin was greatly improved (>90% accuracy for all populations). We also calculated effective size with and without removing physically linked loci identified from a linkage map, a novel method for nonmodel organisms. Estimates of effective size were generally above 1000 and were biased downward when physically linked loci were not removed. Outlier tests based on genetic differentiation identified 733 loci and three genomic regions under putative selection. These markers and genomic regions are excellent candidates for future research and can be used to create high-resolution panels for genetic monitoring and population assignment. This work demonstrates the utility of genomic data to inform conservation in highly exploited species with shallow population structure.

Regions of the genome displaying elevated differentiation (genomic islands of divergence) are thought to play an important role in local adaptation, especially in populations experiencing high gene flow. However, the characteristics of these islands as well as the functional significance of genes located within them remain largely unknown. Here, we used data from thousands of SNPs aligned to a linkage map to investigate genomic islands of divergence in three ecotypes of sockeye salmon (Oncorhynchus nerka) from a single drainage in southwestern Alaska. We found ten islands displaying high differentiation among ecotypes. Conversely, neutral structure observed throughout the rest of the genome was low and not partitioned by ecotype. One island on linkage group So13 was particularly large and contained six SNPs with F > 0.14 (average F of neutral SNPs = 0.01). Functional annotation revealed that the peak of this island contained a nonsynonymous mutation in a gene involved in growth in other species (TULP4). The islands that we discovered were relatively small (80-402 Kb), loci found in islands did not show reduced levels of diversity, and loci in islands displayed slightly elevated linkage disequilibrium. These attributes suggest that the islands discovered here were likely generated by divergence hitchhiking; however, we cannot rule out the possibility that other mechanisms may have produced them. Our results suggest that islands of divergence serve an important role in local adaptation with gene flow and represent a significant advance towards understanding the genetic basis of ecotypic differentiation.

In their recently corrected manuscript, "Breaking RAD: An evaluation of the utility of restriction site associated DNA sequencing for genome scans of adaptation", Lowry et al. argue that genome scans using RADseq will miss many loci under selection due to a combination of sparse marker density and low levels of linkage disequilibrium in most species. We agree that marker density and levels of LD are important considerations when designing a RADseq study; however, we dispute that RAD-based genome scans are as prone to failure as Lowry et al. suggest. Their arguments ignore the flexible nature of RADseq; the availability of different restriction enzymes and capacity for combining restriction enzymes ensures that a well-designed study should be able to generate enough markers for efficient genome coverage. We further believe that simplifying assumptions about linkage disequilibrium in their simulations are invalid in many species. Finally, it is important to note that the alternative methods proposed by Lowry et al. have limitations equal to or greater than RADseq. The wealth of studies with positive impactful findings that have used RAD genome scans instead supports the argument that properly conducted RAD genome scans are an effective method for gaining insight into ecology and evolution, particularly for non-model organisms and those with large or complex genomes.