DGEN: A Test Statistic for Detection of General Introgression Scenarios

When two species hybridize, one outcome is the integration of genetic material from one species into the genome of the other, a process known as introgression. Detecting introgression in genomic data is a very important question in evolutionary biology. However, given that hybridization occurs between closely related species, a compli-cating factor for introgression detection is the presence of incomplete lineage sorting, or ILS. The D -statistic, famously referred to as the “ABBA-BABA” test, was pro-posed for introgression detection in the presence of ILS in data sets that consist of four genomes. More recently, D FOIL —a set of statistics—was introduced to extend the D -statistic to data sets of five genomes. The major contribution of this paper is demonstrating that the invariants underly-ing both the D -statistic and D FOIL can be derived automatically from the probability mass functions of gene tree topologies under the null species tree model and alterna-tive phylogenetic network model. Computational requirements aside, this automatic derivation provides a way to generalize these statistics to data sets of any size and with any scenarios of introgression. We demonstrate the accuracy of the general statistic, which we call D GEN , on simulated data sets with varying rates of introgression, and apply it to an empirical data set of mosquito genomes. We have implemented D GEN and made it available, both as a graphical user interface tool and as a command-line tool, as part of the freely available, open-source software package ALPHA ( https://github.com/chilleo/ALPHA ).