Distinguishing noise from signal in patterns of genomic divergence in a highly polymorphic avian radiation

Abstract

Recently diverged taxa provide the opportunity to search for the genetic basis of the phenotypes that distinguish them. Genomic scans aim to identify loci that are diverged with respect to an otherwise weakly differentiated genetic background. These loci are candidates for being past targets of selection because they behave differently from the rest of the genome that has either not yet differentiated or that may cross species barriers through introgressive hybridization. Here we use a reduced-representation genomic approach to explore divergence among six species of southern capuchino seedeaters, a group of recently radiated sympatric passerine birds in the genus Sporophila. For the first time in these taxa, we discovered a small proportion of markers that appeared differentiated among species. However, when assessing the significance of these signatures of divergence, we found that similar patterns can also be recovered from random grouping of individuals representing different species. A detailed demographic inference indicates that genetic differences among Sporophila species could be the consequence of neutral processes, which include a very large ancestral effective population size that accentuates the effects of incomplete lineage sorting. As these neutral phenomena can generate genomic scan patterns that mimic those of markers involved in speciation and phenotypic differentiation, they highlight the need for caution when ascertaining and interpreting differentiated markers between species, especially when large numbers of markers are surveyed. Our study provides new insights into the demography of the southern capuchino radiation and proposes controls to distinguish signal from noise in similar genomic scans.

Exploring the phylogeography of a hexaploid freshwater fish by RAD sequencing

“…The average GC content of the reads obtained through RAD sequencing, between 38.5% and 40.8%, was similar to that reported for the zebrafish, Danio rerio , genome (38.6%) (Zhou, Bizzaro, & Marx, 2004) giving confidence that these data were not particularly biased by our choice of AT‐rich restriction sites (contrary to Campagna, Gronau, Silveira, Siepel, & Lovette, 2015; DaCosta & Sorenson, 2014). The relatively high transition/transversion ratio of 2.04 after filtering for paralogs and mitochondrial loci may indicate a bias toward genic regions, as SNPs occur more frequently as transitions in exons than in introns (Park, Yu, Mun, & Lee, 2010).…”

“…As the level of missing data is allowed to increase, so does the signal of phylogeographic structure. This phenomenon has been briefly described in the literature (Campagna et al., 2015; Huang & Knowles, 2014; Takahashi, Nagata, & Sota, 2014; Wagner et al., 2013) and should be viewed as a motivation to include more missing data to reduce potential biases from only examining highly conserved regions. However, our final datasets retained relatively low levels of missing data—19.7% within our 723 SNP dataset and 0.63% within the DIYABC reduced SNP dataset of 661 markers.…”

Outlier analyses to test for local adaptation to breeding grounds in a migratory arctic seabird

“…In some cases, the use of a high number of markers can increase resolution: In the black‐footed albatross ( Phoebastria nigripes ) for instance, genotyping of thousands of genomewide markers revealed weak but significant population structure among colonies in Japan and Hawaii, where previous results from traditional genetic markers were contradictory (Dierickx, Shultz, Sato, Hiraoka, & Edwards, 2015). However, similar to our results, in other studies where mtDNA and microsatellite data failed to detect population structure, the use of thousands of markers did not detect genetic differentiation, even among species (e.g., Campagna et al., 2015). …”

“…For example, in southern capuchino seedeaters (genus Sporophila ), this approach revealed that putative differentiated loci were not really informative of differentiation among species because the same pattern could be obtained from a randomized dataset (Campagna et al., 2015). In our case, however, outlier loci identified from the randomized dataset failed to indicate any structuring.…”

“…We created a randomized genotype dataset by randomly reassigning population labels to each of the thick‐billed murre samples as outlined in Campagna, Gronau, Silveira, Siepel, and Lovette (2015). Then, we ran outlier analyses on this dataset using lositan and used outlier loci to investigate population structure in structure .…”