Presentation Detail

SSB Symposium: Ernst Mayr Symposium

Eaton, Deren [1], Ree, Richard [2].

Detecting Genomic Introgression at the Phylogenetic Scale.

Introgression describes the incorporation of genes from one lineage into another. Unlike hybridization which can be directly observed, introgression occurs over multiple generations, leaving a diffuse signal that is only detectable through comparing many genomic markers from across multiple related taxa. Although introgression is sometimes inferred from hybridization, hybrids are not always observable -- occurring either very rarely or only in the past -- and thus introgression may be more common than expected. With the increased availability of genomic sequence data among non-model organisms we now have the opportunity to test for introgression more broadly and thus to better understand its potential impacts on character evolution and phylogenetic inference. Using simulations I evaluated the performance of Patterson's D-statistic -- a recently developed method for detecting introgression, most notably used to infer interbreeding between humans and Neanderthals -- and show that it yields a high rate of false positives when applied at deeper phylogenetic scales and in more diverse clades. The main source of error comes from sampling one or more close relatives of a lineage that hybridized with a more distant taxon, but which themselves did not hybridize: because these lineages share many derived alleles by descent, introgression from one will necessarily introduce alleles shared by the other, causing close relatives to exhibit a similar signal of having introgressed. To overcome this problem I developed a novel extension to this test termed the partitioned D-statistic, which separates introgression from the signal of shared ancestry among multiple candidate taxa, allowing detection of introgression from each lineage separately. In addition to removing a source of error, the partitioned metric is also able to indicate the direction of gene flow, which the original test did not. As an empirical example, I applied the partitioned D-statistic test to RADseq data from a clade of flowering plants (Pedicularis: Orobanchaceae) with no prior evidence of hybridization. Multiple independent introgression events were detected that would have been otherwise indistinguishable using the standard D-statistic, and which were consistent with the geographic overlap of taxa. In light of these results, alternative phylogenetic hypotheses were compared while minimizing the effect of unidirectionally introgressed DNA, recovering support for a topology more consistent with geography and morphology, but which had formerly been obscured by introgression.

Related Links:
D. Eaton -- academic website

1 - University of Chicago, Committee on Evolutionary Biology, 1025 E. 57th St., Chicago, IL, 60637, USA
2 - Field Museum Of Natural History, Botany, 1400 S Lake Shore Drive, Chicago, IL, 60605, USA

Keywords:
plants
RADseq
hybridization
concordance
D-statistic.

Presentation Type: Symposium Presentation
Session: 210
Location: Ballroom 2/Cliff Lodge
Date: Sunday, June 23rd, 2013
Time: 9:15 AM
Number: 210004
Abstract ID:538
Candidate for Awards:Ernst Mayr Award