TwisstNTern 2: Ternary analysis of topology weights from tree sequences

Hila Lifchitz, Sanketh Vedula, Arka Pal, Daria Shipilina, Sean Stankowski

Recent advances in genealogical inference now allow the reconstruction of genome-wide sequences of trees for large sets of samples, providing detailed records of how evolutionary relationships vary along the genome. Tree sequences encode a vast amount of information, but new approaches are needed to extract relevant patterns and make inferences. TwisstNtern is a program for visualising and analysing topology weights from four-population tree sequences. TwisstNtern takes a wide range of tree sequence formats as input, conducts topology weighting using the Twisst algorithm, and projects the topology weights in a ternary plot. This enables intuitive visualisation of the joint distribution of weights, and formal tests for asymmetrical genealogical discordance caused by processes such as introgression. The package also includes tools for simulation, significance testing, and the comparison of distributions between empirical and simulated datasets.