Bayesian analyses in phylogenetic palaeontology: interpreting the posterior sample

Abstract

Establishing hypotheses of relationships is a critical prerequisite for any macroevolutionary analysis, but different approaches exist for achieving this goal. Amongst palaeontologists using morphological data the Bayesian approach is increasingly preferred over parsimony, but this shift also alters the way we think about samples of trees. Here we revisit stratigraphic congruence as a comparator between Bayesian and parsimony samples, but in a new visual context: treespace. Such spaces represent an ordination of unique topologies that can also be extended to create a ‘landscape’ where altitude represents some comparative measure (here congruence with stratigraphy). By co‐opting existing visualization tools and applying them to a meta‐analysis of 128 cladistic data sets we show that there is no consistent favouring of either Bayesian or parsimony according to stratigraphic congruence metrics, and further that empirical treespace visualizations suggest a complex variety of topological landscapes. We conclude by arguing that treespaces should become a standard exploratory tool in phylogenetic analysis.

New opabiniid diversifies the weirdest wonders of the euarthropod lower stem group

“…Such plots identify whether uncertainty in support for opabiniid relationships in the posterior sample ( n = 4512 trees for analyses where proboscis is coded as present; Table 1 ) and MPTs ( n = 12 trees) is restricted to tree islands with otherwise similar topologies, or spread throughout a large region of occupied treespace. While treespace has been previously explored in meta-analyses of fossil datasets (Brazeau, Guillerme and Smith, 2019; Koch and Parry, 2020; Wright and Lloyd, 2020), this is, to our knowledge, the first attempt to use such a visualization to interrogate the distribution of bipartitions for the position of a focal fossil taxon. Several possible hypotheses are subsets: KUMIP 314087 could be part of a clade with either Opabinia or Deuteropoda (pink and dark purple colors, respectively, in Fig.…”

“…Such calculations have been effective in summarizing the taxonomic uncertainty in fossil placement (Klopfstein and Spasojevic, 2019). Furthermore, our visualization of the sample of optimal trees (Hillis, Heath and St. John, 2005; St. John, 2017; Wright and Lloyd, 2020) illustrates the distribution of topological distances between conflicting and overlapping hypotheses. This technique allows the strength of support for competing hypotheses of relationships to be more comprehensively evaluated beyond an arbitrary cutoff value.…”

“…Supplemental to traditional clade support metrics, we used classical multidimensional scaling (MDS) to plot treespace (Gower, 1966; Hillis, Heath and St. John, 2005; St. John, 2017; Wright and Lloyd, 2020), with the goal of identifying the distribution of trees resolving key clades formed with KUMIP 314087 ( Table 1 ). Our R script inputs the unrooted post-burnin posterior samples (resultant from BI) and MPTs (resultant from MP) using ape v.5.3 (Paradis and Schliep, 2019), and employs phangorn v.2.5.5 (Schliep, 2011) to calculate pairwise unweighted Robinson-Foulds distances (RF, the proportion of bipartitions defined by a branch in one tree that is lacking in another tree) (Robinson and Foulds, 1981) for the total set of trees resulting from all analyses.…”

“…KUMIP 314087 is a new genus and species that shares characters with both radiodonts and Opabinia regalis . We evaluate its phylogenetic position using both maximum parsimony (MP) and Bayesian inference (BI) and further interrogate the support for alternative relationships for KUMIP 314087 by visualizing the frequency and variation of these alternatives in treespace (Hillis, Heath and St. John, 2005; Wright and Lloyd, 2020). All analyses support an opabiniid affinity for KUMIP 314087.…”

Does time matter in phylogeny? A perspective from the fossil record

“…This stratophenetic vs. cladistics debate was intensely pursued in previous decades [4][5][6][7]. Some argue that temporal evidence provides a valuable source of evidence for distinguishing between competing phylogenetic hypotheses [8][9][10][11][12][13] or for testing different phylogenetic methods [14,15]. Others have argued that incompleteness of the fossil record hinders the reliability of stratophenetics or stratolikelihood, and that morphology alone provides a less biased source of evidence for phylogenetic reconstruction [4].…”