Testing the Utility of Alternative Metrics of Branch Support to Address the Ancient Evolutionary Radiation of Tunas, Stromateoids, and Allies (Teleostei: Pelagiaria)

Abstract

The use of high-throughput sequencing technologies to produce genome-scale datasets was expected to settle some long-standing controversies across the Tree of Life, particularly in areas where short branches occur at deep timescales. Instead, these datasets have often yielded many well-supported but conflicting topologies, and highly variable gene-tree distributions. A variety of branch-support metrics beyond the nonparametric bootstrap are now available to assess how robust a phylogenetic hypothesis may be, as well as new methods to quantify gene-tree discordance. We applied multiple branch support metrics to an ancient group of marine fishes (Teleostei: Pelagiaria) whose interfamilial relationships have proven difficult to resolve due to a rapid accumulation of lineages very early in its history. We analyzed hundreds of loci including published UCE data and newly generated exonic data along with their flanking regions to represent all 16 extant families for more than 150 out of 284 valid species in the group. Branch support was lower for interfamilial relationships (except the SH-like aLRT and aBayes methods) regardless of the type of marker used. Several nodes that were highly supported with bootstrap had very low site and gene-tree concordance, revealing underlying conflict. Despite this conflict, we were able to identify four consistent interfamilial clades, each comprised of two or three families. Combining exons with their flanking regions also produced increased branch lengths in the deep branches of the pelagiarian tree. Our results demonstrate the limitations of employing current metrics of branch support and species-tree estimation when assessing the confidence of ancient evolutionary radiations and emphasize the necessity to embrace alternative measurements to explore phylogenetic uncertainty and discordance in phylogenomic datasets.

Comparing Ultraconserved Elements and Exons for Phylogenomic Analyses of Middle American Cichlids: When Data Agree to Disagree

“…It must also be considered that a single tree most likely will not be capable of encapsulating the complex evolutionary histories of many organisms. Therefore, we agree with previous authors proposing that comparative studies should be based on multiple trees—inferred using different methods and/or molecular markers—to account for uncertainties in the Tree of Life ( Arcila et al 2021 ).…”

“…Another major difference between genomic marker types was the greater gene tree heterogeneity and gene tree-species tree heterogeneity of UCEs compared with exons ( fig. 2 B ), a finding that has also been highlighted in other comparative studies ( Arcila et al 2021 ). This result is not unexpected if we assume that the shorter and less informative UCEs will produce more inaccurate gene trees than the exons ( Camargo et al 2012 ).…”

“…Furthermore, recent discussions have highlighted the fact that nodal support measures, such as nonparametric bootstrap, increase with the size of the data set and can give a false sense of “confidence” in phylogenetic relationships ( Liu et al 2015 ; Simon 2020 ). Therefore, it is necessary to interrogate genomic data sets and explore alternative measures of support to ensure a more rigorous evaluation and a more robust interpretation of the data ( Arcila et al 2021 ). The Nandopsis clade is a phylogenetically difficult lineage to unambiguously resolve, and we are unable to do so comparing two independent and large genomic data sets with robust taxon sampling and varied phylogenetic analyses.…”

“…Recently, it has become increasingly common to interrogate the data used in phylogenetic reconstruction to explore whether data sets contain sufficient unambiguous information to accurately infer species trees ( Doyle et al 2015 ; Arcila et al 2017 ; Alda et al 2019 ; Burbrink et al 2020 ). Yet, most of these studies only explore a single data set, subsets of that data set, or only one type of genomic marker (but see Gilbert et al 2015 ; Reddy et al 2017 ; Betancur-R et al 2019 ; Bossert et al 2019 ; Karin et al 2020 ; Arcila et al 2021 ).…”

Molecular phylogenies map to biogeography better than morphological ones

“…While it is true that morphological trees tend to be less resolved, comparisons restricted to fully resolved trees have demonstrated that real incongruence in their primary phylogenetic signals 59 must account for the differing fits of morphological and molecular trees to biogeography. What we are unable to investigate further without access to the original data and comparative branch support metrics 60 is whether this incongruence is primarily due to lack of information or misleading information in morphological data. If, for example, incongruent relationships in morphological trees are less well supported by indices such as bootstrap 61 or Bremer support 62 than relationships which are congruent with biogeography, it would suggest that the biogeographic incongruence of morphological trees is partly attributable to a lack of strong signal in the morphological data.…”