Updated: Mar 6
For Mar. 4th discussion (led by Monty Slatkin), we revisited Slatkin & Pollack (2008) and Durand et al. (2011) along with the new introgression detection method QuIBL (Eldman et al. 2019), which was designed to delineate introgression from ILS. Despite the intelligent idea and implementation, QuIBL did not consider ancestral subdivisions that could be relevant to the rapid radiation history of Heliconius.
Ancestral subdivisions should be considered as one of the alternatives for introgression detection. If the ancestral population was panmictic, the non-concordant gene trees should exhibit equal frequencies. However, if a subset of the ancestral population persistently (relative to the divergence) interbreed within the subset more than with the rest (due to isolation-by-distance, habitat fragments, or any partial migration barriers) (see illustration below from Durand et al. 2011), this could lead to asymmetries in the nonconcordant gene tree, which could result in “introgression-like” pattern in D and fd statistics in variations of ABBA-BABA tests.
Ancestral subdivisions would also lead to a different distribution of internal branch length (that QuIBL focuses on), so that it is no longer an exponential (rate parameter=1) distribution. QuIBL could rule out ILS when the distribution of internal branch length is not exponential (rate parameter=1), however it could not confidently confirm ‘introgression’ just yet until the part of deviation driven by ancestral subdivision was considered (at least).
Edelman N. B. et al. 2019. Genomic architecture and introgression shape a butterfly radiation. Science. 366: 594-599.
Durand E. Y. et al. 2011.Testing for Ancient Admixture between Closely Related Populations. 28:2239–2252.
Slatkin M. & Pollack J.L. 2008. Subdivision in an Ancestral Species Creates Asymmetry in Gene Tree. Mol. Biol. Evol. 25:2241–2246.