TriLoNet: Piecing Together Small Networks to Reconstruct Reticulate Evolutionary Histories

Abstract

Phylogenetic networks are a generalization of evolutionary trees that can be used to represent reticulate processes such as hybridization and recombination. Here, we introduce a new approach called TriLoNet (Trinet Level- one Network algorithm) to construct such networks directly from sequence alignments which works by piecing together smaller phylogenetic networks. More specifically, using a bottom up approach similar to Neighbor-Joining, TriLoNet constructs level-1 networks (networks that are somewhat more general than trees) from smaller level-1 networks on three taxa. In simulations, we show that TriLoNet compares well with Lev1athan, a method for reconstructing level-1 networks from three-leaved trees. In particular, in simulations we find that Lev1athan tends to generate networks that overestimate the number of reticulate events as compared with those generated by TriLoNet. We also illustrate TriLoNet's applicability using simulated and real sequence data involving recombination, demonstrating that it has the potential to reconstruct informative reticulate evolutionary histories. TriLoNet has been implemented in JAVA and is freely available at https://www.uea.ac.uk/computing/TriLoNet.

Reversals in complex traits uncovered as reticulation events: Lessons from the evolution of parity‐mode, chromosome morphology, and maternal resource transfer

“…These clusters are congruent with the lineages detected by the Bayesian phylogeny (Figure a). The level‐1 network further unraveled the following three reticulation events (Figure b) (a) the entire clade F resulted from reticulation between a parent belonging to a subtype of the viviparous clade D and a parent belonging to a subtype of the oviparous clade A; reticulation existed; (b) within clade A (green cluster); and (c) within the Ibero‐Pyrenean clades (note: excluding the reticulated species B1_09_503; Oldman et al, led to two clades within the violet Iberio‐Pyreanean clade, consisting of a cluster containing B2 and a cluster containing B3 and B1; which is congruent with the Bayesian phylogeny; see also Text S1). The network thus suggests that viviparity evolved once and that clade B might be a reversal.…”

“…Consequently, two possible scenarios of parity mode evolution exist: (a) a single evolution of viviparity (in the ancestor of Anableps and the violet cluster in Figure a) and two potential reversals from viviparity to oviparity, in the ancestor of Tomeurus gracilis and in the ancestor of Oxyzygonectes dovii , or (b) multiple evolution of viviparity (i.e., at least three times, in the ancestor of Jenynsia and Anableps, in the ancestor of Xenodexia, and in the ancestor of the other viviparous species—the ancestor of the violet cluster in Figure a). Testing for reticulation using TriLoNet (Oldman et al, : see Text S1 for detailed methodological details) revealed three clusters in the level‐1 network (Figure b), the gray cluster including Cyprinodon and Jordanella, the green cluster including Fluviphylax, Aplocheilichthys, Oxyzyogonectes, Jenynsia, Anableps, and Xenodexia; and the violet cluster including the remaining species (Figure b). These clusters coincided with the clusters of the Bayesian phylogeny (Figure a).…”

“…Several runs, including 20 and 50 million Markov chain Monte Carlo repetitions were needed in the case of Cyprinodontiformes and Z. vivipara , respectively, for ensuring convergence, which was tested with Tracer v.1.6 (http://tree.bio.ed.ac.uk/software/tracer). In both cases, reticulation was tested using TriLoNet (Oldman, Wu, van Iersel, & Moulton, ), which infers reticulated events from constructing rooted phylogenetic level‐1 networks from sequence alignments (Text S1). TriLoNet employs trinets and it compares well with Lev1athan (Huber, Moulton, & Wu, ), which tends to “overestimate the number of reticulated events as compared with those generated by TriLoNet” (Oldman et al, ).…”

“…TriLoNet employs trinets and it compares well with Lev1athan (Huber, Moulton, & Wu, 2016), which tends to "overestimate the number of reticulated events as compared with those generated by TriLoNet" (Oldman et al, 2016).…”

Binets: Fundamental Building Blocks for Phylogenetic Networks

“…2016). In particular, for a set X of species, a network is constructed for every subset of X size 3 (called a trinet ), and then the trinets are puzzled together to build a network (see Fig.…”