Data from: Exploring rates of change and modes of evolution in blastozoan echinoderms

Data files

Nov 20, 2025 version files 5.05 MB

Gradstein_2020_Augmented.RData

809.46 KB
README.md

8.27 KB
Rock_Unit_Database.RData

4.23 MB

Nov 20, 2025 version files 5.05 MB

Gradstein_2020_Augmented.RData

809.46 KB
README.md

8.67 KB
Rock_Unit_Database.RData

4.23 MB

Abstract

Over the past half-century, paleobiologists have advanced the estimation of phylogenetic relationships among fossil taxa to explore evolutionary patterns in deep time. This study employs phylogenetics, divergence time calculations, and character rate evolution within three blastozoan echinoderm clades: Diploporita, Eublastoidea, and Paracrinoidea. Utilizing Reversible Jump Markov Chain Monte Carlo (rjMCMC) and Fossilized Birth-Death (FBD) models, we investigated evolutionary rates through anatomical subunit partitioning. Results suggest similar rates among the three groups, though Paracrinoidea exhibits elevated rates in several anatomical subunits. The inferred trees largely agree with other recently published analyses and highlight the need to revise echinoderm taxonomy. We tested different clock models for each group and found that model choice had strong effects on resulting trees; our findings suggest that linked clocks had more support than unlinked clocks. These findings indicate a need to carefully consider model choice and rates of evolution when conducting these types of analyses.

https://doi.org/10.5061/dryad.s7h44j1hg

Description of the data and file structure

Models of Character Evolution

A core goal in this study is to examine if the reversible jump Markov Chain Monte Carlo (rjMCMC) algorithm can be used to efficiently compare different hypotheses about character evolution. Reversible Jump Markov Chain Monte Carlo (rjMCMC) is an adaptation of the MCMC algorithm that allows for the number of parameters to change throughout an analysis. In this type of analysis, the model itself is a random variable. For example, in Table S1, you will note that some skyline models have more categories than others. In the case of adding a category, this means (minimally) adding three additional parameters — a new speciation, extinction, and fossil sampling rate. In a traditional MCMC, the number of parameters cannot be varied in an analysis. Thus, in an MCMC, a model must be chosen before running a phylogenetic estimation. In the case of rjMCMC, the model can be evaluated alongside the model parameters, effectively performing model selection at the same time as tree estimation.

Following the termination of each run, R was used (v4.1.2; R Core Team 2022) to summarize the rjMCMC. The proportion of samples spent with a given model is a proxy for its posterior probability. Traditional model selection tools, such as the Bayes Factor, evaluated against the Kass and Raftery (1995) significance scale, can be used to establish if any one model is significantly better than another. A normal MCMC analysis is run until the analysis reaches what is termed convergence. This refers to a time when the MCMC has found an optimal solution and is simply sampling small variations around it. In an rjMCMC estimation, convergence for both the model and the model parameters is needed. Based on our preliminary explorations, this takes longer than simple convergence for model parameters. As such each estimation was run for approximately 1 million generations in RevBayes, with some estimations finishingearlier. We summarized each phylogenetic tree using a maximum clade credibility tree, and summarized parameter distributions in R (Tribble et al. 2022).

Partitioning by anatomical subregion is a common practice in morphological phylogenetics (Clark and Middleton 2008; Close et al. 2015; Simões and Pierce 2021; Casali et al. 2023), as is applying different models of character evolution to different characters (e.g. Hooker 2014). Fortuitously, this simultaneously tests relevant pre- existing hypotheses about evolutionary flexibility among different parts of organisms that are generated by functional biological theory and evolutionary-developmental theory (Wright et al. 2021). We use the novel rjMCMC approach to evaluate support between five different character models. These models include a basic Mk model (Lewis 2001), in which all characters are evolving under the same set of assumptions. The other four models all involve some anatomical subunit partitioning as described above.

Aside from the character partitions outlined on Table 3, all phylogenetic models also used among-character rate variation (Yang 1994) to allow for individual characters having different rates of evolution without assuming a particular “high” or “low” rate for each character a priori. In one set of analyses, rates of evolution via the clock model were linked across partitions. In a second set of estimations, the rates of morphological evolution were unlinked between partitions, thus allowing each module of characters to have its own evolutionary rate.

Phylogenetic Analyses

Fossilized birth-death (FBD) analyses require upper and lower bounds on the possible first appearances of analyzed taxa in order to calculate the prior probabilities on phylogenies (see below). For example, a blastozoan taxon first appearing within the *Belodina compressa *conodont zone of North America might have a first appearance ranging from 455.2 – 454.2 Ma (Gradstein et al. 2020). We use data from the Paleobiology Database (PBDB) to identify the oldest possible occurrences. Because the ages provided by the PBDB are fairly general, we use a separate dataset (see Congreve et al. 2021) to provide more exact ages based on conodont, graptolite and ammonoid biozonations and the Gradstein and others (2020) time scale. The stratigraphic data providing the lower and upper bounds of first appearance dates reflects 312 blastozoan occurrences from the Paleobiology Database, which come from 241 localities and 117 rock units. Taxonomic data for all relevant species reflect currently accepted opinions given 274 entered opinions, including the opinions of the relevant authors of this paper. A total of 178 published works contributed to these data, the most important of which include: Frest and others (1999), Regnéll (1945), Parsley and Mintz (1975), Fay (1961), Bassler and Moodey (1943), Macurda (1983), Fay and others (1982), Wilson (1946), Krivicich and others (2014), Paul (1971), and Bauer (2018).

We performed our phylogenetic analyses in RevBayes (version 1.2.1, Höhna et al. 2016). We estimated dated phylogenetic trees using the Fossilized Birth-Death model (FBD; Heath 2014). Under the FBD, it is assumed that the tree can be then described by four core parameters: speciation, extinction, and sampling (fossil or present). We identified reasonable priors on these parameters using values from Wright and others (2021). We also tested if a time-homogeneous FBD process was a good descriptor of our data, or if a skyline model would be appropriate. In a skyline model, the rates of speciation, extinction and sampling may change between time intervals. This may be expected if different geological periods present different selective pressures.

Files and variables

File: Raw_Data-2024.zip

Description: Raw morphological character matrices and fossil ages. These are generated for use with the phylogenetic software RevBayes 1.2.1. Character matrices are in NEXUS file format, and can be opened to view in a text editor. Fossil age files, containing a column for each taxon name, minimum and maximum ages, are in CSV format and can be viewed in a text editor or spreadsheet views.

File: Final_Output.zip

Description: Trees and tree figures estimated in phylogenetic analyses using RevBayes 1.2.1. These are in NEXUS (.tre) format.

File: Intermediate_Output.zip

Description: MCMC log files generated by RevBayes. These were summarized into the Figure 3-7 in Final_Output using RevBayes, RStudio, and RevGadgets. These files are in .log format, and can be viewed in Tracer, or opened as tables in R/RStudio.

Gradstein_2020_Augmented.RData

Description: R data file of geologic sampling. This can be opened in R or R Studio. Contains raw stratigraphic data that can be opened into a dataframe-like object, and used to precisely calculate fossil ranges.

Rock_Unit_Database.RData

Description: R data file of geologic timescale. This can be opened in R or R Studio. Will expand to a dataframe object containing columns of taxonomic occurrences, and minimum and maximum ages in millions of years.

Supplemental_Sheffield_et_al._Alternative_Text_for_Figures.docx

Description: A Microsoft Word document containing alternative text for readers using assistive visual technology.

Supplementary_Tables_and_Figures.docx

Description: A Microsoft Word document containing Table S2, and Figures S1-3.

Table_S1.xlsx

Description: A Microsoft Excel document containing Table S1.

Code/software

Scripts directory:

RevBayes analytical scripts
- For each taxon (Blastoidea, Paracrinoidea, Diploplorans), there is a directory. In that directory are the .Rev scripts (for use with RevBayes minimally v1.2.1). The main script, rjFBD.Rev loads the competing sequence model files in that folder. The main script, rjFBD.Rev is what should be executed to reproduce the paper.
The R scripts directory: Contains two files for output processing. The RevBayes analytical scripts must be run before these can be used.
- PrettyTrees.R: For generating the tree figures from the .tre files. Can be used with R or RStudio.
- BFCalc.RmD. For generating Figures 3 and 4, using the posterior traces resulting from the RevBayes analytical scripts. Can be used with R or RStudio.

Access information

Data was derived from the following sources:

PBDB. Clones of the database snapshots used in this paper are provided as .Rdata files

Recent work suggests variation in rates of evolution of echinoderm traits (Novack-Gottshall et al. 2022). In prior work, it has been hypothesized that “functional” traits like feeding characters may evolve at faster rates than other traits, a hypothesis not limited to echinoderms (e.g., Foote 1995; Ciampaglio 2002; Smith and Hopkins 2015; Wright 2017), but seen in animals in general (Wagner 1995; Sánchez-Villagra and Williams 1998). The morphological diversity seen in blastozoan echinoderms has frequently hindered our abilities to establish homology amongst characters, let alone examine rates of evolution. However, with the development of echinoderm-specific homology schemes (Mooi and David 1997; Sumrall 2010) and the application of these schemes to many blastozoan groups, these questions of evolution questions can be addressed. Both homology schemes have strengths and weaknesses (Sumrall et al. 2023), but, because this particular study is focused on Paleozoic blastozoans, and the modeling of morphological shifts of the oral and ambulacral systems, the morphological character datasets utilized in this particular study (i.e. Sheffield and Sumrall 2018; Bauer 2020; Limbeck et al. 2024) were developed predominantly using UEH with some influence from EAT. We explored anatomical subunit partitioning in this study using three groups of blastozoans, each of which is variable in its temporal and geographic patterns and in its body plans (Table 1). Each of these groups — Eublastoidea (Bauer 2021), Diploporita (Sheffield and Sumrall 2019a), and Paracrinoidea (Limbeck et al. 2024) — has recently published phylogenetic analyses from which we have pulled character data to build the models tested herein.

Character groups were used to partition the data in downstream analyses. The three character matrixes were compiled separately for different studies so for this work, the types of characters were considered and grouped based on the following seven anatomical features: (1) Respiratory: characters that are describing part of the respiratory structures; (2) Oral Plate Circlet: characters that include the plates from the oral plate circlet series as defined in Sumrall and Waters (2012); (3) Feeding: characters related to the ambulacral and brachiolar structures; (4) Periproct Area: characters related to the anus opening and plates of the anal area; (5) Reproduction: characters related to the gonopore (i.e. the genital pore); (6) Theca: characters that broadly represent the thecal body and do not fall into the other assigned character groups; and (7) Attachment structure: characters related to the stem or holdfast (Table 2; refer to Fig. 1 for examples of these characters for each set).