“…In addition to identifying sweeps from single large values of G12 and G123, we find that the genomic signature of these MLG-based statistics surrounding the site of selection provides a means of distinguishing a sweep from other types of selection (e.g., balancing selection). This additional layer of differentiation motivates the use of MLG identity statistics as a signature in a statistical learning framework, as such approaches have increasing in prominence for genome analysis [Grossman et al, 2010, Lin et al, 2011, Pavlidis et al, 2010, Ronen et al, 2013, Pybus et al, 2015, Ronen et al, 2015, Sheehan and Song, 2016, Schrider and Kern, 2016, Akbari et al, 2017, Kern and Schrider, 2018, Mughal and DeGiorgio, 2018]. We expect that the MLG-based approaches G12 and G123, in conjunction with G2/G1, will be invaluable in localizing and classifying adaptive targets in both model and non-model study systems.…”