CoChat
Sign Up
Cited by 292Marine Invertebrate Physiology and EcologyGenomics and Phylogenetic StudiesProtist diversity and phylogeny

Improved Modeling of Compositional Heterogeneity Supports Sponges as Sister to All Other Animals

Open access
Feuda, Roberto, Dohrmann, Martin, Pett, Walker, Philippe, Hervé, Rota-Stabelli, Omar, Lartillot, Nicolas, Wörheide, Gert, Pisani, Davide·2017·Current Biology
DOI 10.1016/j.cub.2017.11.008
Open this paper in CoChat — free accountOpen PDF

Abstract

The relationships at the root of the animal tree have proven difficult to resolve, with the current debate focusing on whether sponges (phylum Porifera) or comb jellies (phylum Ctenophora) are the sister group of all other animals [1-5]. The choice of evolutionary models seems to be at the core of the problem because Porifera tends to emerge as the sister group of all other animals ("Porifera-sister") when site-specific amino acid differences are modeled (e.g., [6, 7]), whereas Ctenophora emerges as the sister group of all other animals ("Ctenophora-sister") when they are ignored (e.g., [8-11]). We show that two key phylogenomic datasets that previously supported Ctenophora-sister [10, 12] display strong heterogeneity in amino acid composition across sites and taxa and that no routinely used evolutionary model can adequately describe both forms of heterogeneity. We show that data-recoding methods [13-15] reduce compositional heterogeneity in these datasets and that models accommodating site-specific amino acid preferences can better describe the recoded datasets. Increased model adequacy is associated with significant topological changes in support of Porifera-sister. Because adequate modeling of the evolutionary process that generated the data is fundamental to recovering an accurate phylogeny [16-20], our results strongly support sponges as the sister group of all other animals and provide further evidence that Ctenophora-sister represents a tree reconstruction artifact. VIDEO ABSTRACT.

Verified 2026-10-06 · 5 references not checked yet

AI snapshot

Written from the paper’s full text.

TL;DR

Feuda et al. use posterior predictive tests to show that two phylogenomic datasets previously supporting Ctenophora-sister are poorly described by standard models, especially for site-specific and lineage-specific compositional heterogeneity. When data are recoded and analyzed with CAT-GTR+G, model adequacy improves and Porifera-sister is favored, leading the authors to conclude that Ctenophora-sister is a tree reconstruction artifact.

Key findings

  • All considered models were rejected for site-specific and lineage-specific compositional heterogeneity, but CAT-GTR+G performed best. For WhelanD20-Opistho, PPA-DIV Z scores were 127.35 (WAG), 116.17 (LG), 92.62 (GTR), 104.71 (PF-scheme), and 6.21 (CAT-GTR); for Chang, CAT-GTR gave 6.17. Recoding reduced these values, e.g., Dayhoff-6 + CAT-GTR gave 0.94 for WhelanD20-Opistho and KGB-6 + CAT-GTR gave 0.71 for Chang. For PPA-MAX, all models were strongly rejected, with the smallest absolute Z score being 12.15 (CAT-GTR on Chang).
  • WAG+G, LG+G, GTR+G (with or without recoding), and PF-schemes invariably supported Ctenophora-sister (posterior probabilities [PPs] ≥ 1). CAT-GTR+G on amino acid datasets varied by outgroup: WhelanD20-Opistho supported Ctenophora-sister (PP ≥ 1), whereas WhelanD20-Holo and WhelanD20-Choano supported Porifera-sister (PP = 0.68 and 0.77, respectively); Chang under CAT-GTR+G supported Ctenophora-sister (PP ≥ 1).
  • CAT-GTR+G analyses of recoded datasets increased support for Porifera-sister. Dayhoff-6 PPs for Porifera-sister were 0.89 (WhelanD20-Opistho), 0.99 (WhelanD20-Holo), 0.95 (WhelanD20-Choano), and 0.74 (Chang); S&R-6 PPs were 0.98, 0.99, 0.94, and 0.62; KGB-6 PPs were 0.42, 0.68, 0.37, and 0.78. Ctenophora-sister received lower support in these analyses.
  • Additional tests rejected the claim that Porifera-sister signal is limited to outlier genes: a Shen et al. outlier-excluded dataset still supported Porifera-sister under CAT-GTR+G (PP = 0.98 and 0.99 with and without Dayhoff-6 recoding), and a Dayhoff-6 recoded version of the Whelan et al. 2017 dataset supported Porifera-sister with PP = 1.

Methods

  • Posterior predictive analyses (PPAs) tested model adequacy on WhelanD20 (70 taxa) and Chang (77 taxa, 51,940 amino acid positions) using WAG+G, LG+G, GTR+G, optimal PartitionFinder schemes, and CAT-GTR+G. Site-specific amino acid preferences were assessed with PPA-DIV, PPA-CONV, and PPA-VAR; lineage-specific compositional heterogeneity was assessed with PPA-MAX and PPA-MEAN, with results quantified as Z scores.
  • Data recoding used Dayhoff-6, S&R-6, and KGB-6. Phylogenetic analyses were performed in PhyloBayes MPI v1.7a for GTR+G and CAT-GTR+G (with and without recoding), and RAxML v8.2.9 was used for the Chang PF-scheme analysis; convergence was assessed with bpcomp and tracecomp.
  • Additional analyses applied CAT-GTR+G to a Shen et al. outlier-excluded dataset with and without Dayhoff-6 recoding, and to a Dayhoff-6 recoded version of the Whelan et al. 2017 dataset.

Limitations

  • Stated: none of the considered models explicitly accounts for lineage-specific compositional heterogeneity, and all models—including CAT-GTR+G—were strongly rejected in PPA-MAX and PPA-MEAN tests (smallest absolute Z = 12.15). Models that jointly account for site and lineage heterogeneity are noted as computationally too demanding for phylogenomic datasets.
  • Stated: recoding can cause signal erosion, although the authors argue this would collapse nodes rather than change the supported topology; also, the CAT-GTR+G analysis of the amino acid version of Chang did not converge well, while all other analyses did.
  • Evident: the study’s conclusions depend on a selected set of models and PPA statistics; CAT+G was not directly tested (CAT-GTR+G was used as a representative), and the PartitionFinder protocol excluded some available models such as LG4X.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1016/j.cub.2017.11.008

Cites

67 of 72 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 109 cited sources checked · 0 supported · 0 flagged

  • Findingown data
    § SUMMARY · p. 3

    No routinely used evolutionary model can adequately describe both site-specific and lineage-specific compositional heterogeneity in the WhelanD20 and Chang datasets.

    “no routinely used evolutionary model can adequately describe both forms of heterogeneity”
    Quote in the paperSubstantial claim
  • Findingown data
    § RESULTS AND DISCUSSION · p. 3

    Models previously used to study animal phylogeny (WAG+G, LG+G, GTR+G, PF-schemes, CAT-GTR+G) fell short of adequately describing among-site amino acid preferences in both WhelanD20 and Chang datasets.

    “We showed that for both datasets, models that have previously been used to study the animal phylogeny fell short, to various degrees, of adequately describing among-site amino acid preferences”
    • Also stated in: § SUMMARY · p. 4 · plain
      “showing that all models fell short of adequately describing site-specific amino acid preferences”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [21] Whelan, 2001, A general empirical model of protein evolution derived from multiple protein families using a maximum-likelihood approach, Mol. Biol. Evol. source

      Not checked in the free run.

    • cites [22] Le, 2008, An improved general amino acid replacement matrix, Mol. Biol. Evol. source

      Not checked in the free run.

    • cites [23] Yang, 1994, Estimating the pattern of nucleotide substitution, J. Mol. Evol. source

      Not checked in the free run.

  • Findingown data
    § SUMMARY · p. 4

    PPA-DIV is more discriminatory than other PPA tests, having a broader distribution of Z scores across models.

    “A comparison of the spread of the Z scores obtained, across all models and from all of our PPAs (Tables 1 and S1), indicates that PPA-DIV is more discriminatory: it has a broader distribution of Z scores across models.”
    Quote in the paperNot a substantial claim
  • Findingown data
    § Improved Modeling of Heterogeneity Supports PoriferaSister · p. 5

    CAT-GTR+G analyses of WhelanD20 recovered either Ctenophora-sister or Porifera-sister depending on the outgroups used.

    “Differently, CAT-GTR+G analyses of WhelanD20 recovered either Ctenophora-sister or Porifera-sister (Porifera as the sister group of all other animals) depending on the outgroups used.”
    Quote in the paperSubstantial claim
  • Findingcitation
    § SUMMARY · p. 3

    Data-recoding methods reduce compositional heterogeneity in the WhelanD20 and Chang datasets.

    “We show that data-recoding methods [13-15] reduce compositional heterogeneity in these datasets”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [13] Hrdy, 2004, Trichomonas hydrogenosomes contain the NADH dehydrogenase module of mitochondrial complex I, Nature source

      Not checked in the free run.

    • cites [14] Susko, 2007, On reduced amino acid alphabets for phylogenetic inference, Mol. Biol. Evol. source

      Not checked in the free run.

    • cites [15] Kosiol, 2004, A new criterion and method for amino acid classification, J. Theor. Biol. source

      Not checked in the free run.

  • May combine two claims. 25
  • Classified as probably not a substantial claim. 25

  • May combine two claims.

    Findingown data
    § Improved Modeling of Heterogeneity Supports PoriferaSister · p. 5

    WhelanD20-Holo and WhelanD20-Choano supported Porifera-sister (PP = 0.68 and PP = 0.77, respectively).

    “In contrast, WhelanD20-Holo, the dataset from which fungal outgroups were excluded, and WhelanD20-Choano, the dataset from which all outgroups but the Choanoflagellata were excluded, supported Porifera-sister (PP = 0.68 and PP = 0.77, respectively; Table 3; Figures S2B and S3B).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § SUMMARY · p. 4

    All models, including CAT-GTR+G, were strongly rejected for lineage-specific compositional heterogeneity, with the smallest absolute Z score (|Z| = 12.15) from CAT-GTR+G analysis of Chang.

    “However, all models, including CAT-GTR+G, were strongly rejected-with the smallest absolute Z score ( j Z j = 12.15) obtained from the CAT-GTR+G analysis of Chang.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § SUMMARY · p. 4

    With PPA-MAX and PPA-MEAN tests, CAT-GTR+G obtained the lowest Z scores, and PF-schemes emerged as the second best modeling strategy.

    “With both tests, CAT-GTR+G obtained the lowest Z scores, and PF-schemes emerged as the second best modeling strategy”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § Improved Modeling of Heterogeneity Supports PoriferaSister · p. 5

    Support for Ctenophora-sister decreases, while support for Porifera-sister increases, when heterogeneity in the data is better accounted for.

    “Our results show that support for Ctenophora-sister decreases, while at the same time support for Porifera-sister increases, when heterogeneity in the data is better accounted for.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § SUMMARY · p. 4

    All recoded datasets obtained lower Z scores than the original amino acid datasets for both PPA-MAX and PPA-MEAN, indicating that recoding is effective at reducing lineage-specific compositional heterogeneity.

    “All recoded datasets obtained lower Z scores than the original amino acid datasets for both PPA-MAX and PPAMEAN (Tables 1 and S1), indicating that recoding is effective at reducing lineage-specific compositional heterogeneity.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Improved Modeling of Heterogeneity Supports PoriferaSister · p. 5

    For KGB-6, the PP for Porifera-sister was 0.78 for Chang and 0.68 for WhelanD20-Holo.

    “For KGB-6, the PP for Porifera-sister was 0.78 for Chang and 0.68 for WhelanD20-Holo.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § SUMMARY · p. 5

    Under CAT-GTR+G, only PPA-DIV invariably recovered reduced Z scores from the recoded datasets, with Z scores from PPA-CONV and PPA-VAR being similar in analyses with and without recoding.

    “Under CAT-GTR+G, only PPA-DIV invariably recovered reduced Z scores from the recoded datasets (Table 1), with Z scores from PPA-CONV and PPA-VAR being similar in analyses performed with and without recoding (Table S1).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Improved Modeling of Heterogeneity Supports PoriferaSister · p. 5

    Phylogenies inferred from Chang using PF-scheme and CAT-GTR+G supported Ctenophora-sister with 100% bootstrap support and PP ≥ 1, respectively.

    “Similarly, phylogenies inferred from Chang using PF-scheme and CAT-GTR+G (see STAR Methods) supported Ctenophora-sister with 100% bootstrap support and PP z 1, respectively”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § Improved Modeling of Heterogeneity Supports PoriferaSister · p. 5

    Support for Ctenophora-sister was severely reduced in CAT-GTR+G analyses of both recoded datasets, with support for Porifera-sister being higher than that for Ctenophora-sister.

    “Support for Ctenophora-sister was severely reduced in CATGTR+Ganalyses of both recoded datasets, with support for Porifera-sister being higher than that for Ctenophora-sister”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Weighing the Evidence · p. 5

    PF-schemes are much worse than CAT-GTR+G and are not always better than unpartitioned modeling strategies at capturing the heterogeneity of the considered datasets.

    “More broadly, our results demonstrate that PFschemes (strongly advocated for by [10, 28, 29, 38]) are much worse than CAT-GTR+G and are not always better than unpartitioned modeling strategies at capturing the heterogeneity of the considered datasets”
    Quote in the paperSubstantial claim0 of 4 sources checked
    • cites [10] Whelan, 2015, Error, signal, and the placement of Ctenophora sister to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

    • cites [28] Halanych, 2016, Miscues misplace sponges, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

    • cites [29] Whelan, 2017, Who let the CAT out of the bag? Accurately dealing with substitutional heterogeneity in phylogenomic analyses, Syst. Biol.

      Not checked in the free run.

    • cites [38] Whelan, 2017, Ctenophore relationships and their placement as the sister group to all other animals, Nat. Ecol. Evol. source

      Not checked in the free run.

  • May combine two claims.

    Findingown data
    § Weighing the Evidence · p. 6

    WAG+G, LG+G, GTR+G (with or without recoding), and PF-schemes describe the data worse than CAT-GTR+G and invariably found support for Ctenophora-sister.

    “WAG+G, LG+G, GTR+G (with or without recoding), and PF-schemes describe the data worse than CAT-GTR+G and invariably found support for Ctenophora-sister.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Weighing the Evidence · p. 6

    A CAT-GTR+G analysis of the Dayhoff-6 recoded version of the dataset of Whelan et al. [38] reached excellent convergence and found strong support for Porifera-sister (PP = 1).

    “Indeed, a CAT-GTR+G analysis of the Dayhoff-6 recoded version of this dataset reached excellent convergence and found strong support for Porifera-sister (PP = 1; Figure S3K)”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [38] Whelan, 2017, Ctenophore relationships and their placement as the sister group to all other animals, Nat. Ecol. Evol. source

      Not checked in the free run.

  • May combine two claims.

    Findingown data
    § Weighing the Evidence · p. 6

    Data recoding reduced the impact of heterogeneity on phylogenetic analyses and lent support to Porifera-sister, independently of the outgroups included.

    “Data recoding reduced the impact of heterogeneity on our phylogenetic analyses (see also [7]), and this lent support to Porifera-sister. Notably, this result emerged independently of the outgroups included.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [7] Simion, 2017, A large and consistent phylogenomic dataset supports sponges as the sister group to all other animals, Curr. Biol. source

      Not checked in the free run.

  • May combine two claims.

    Findingown data
    § Figure 1. CAT-GTR+G Trees from Dayhoff-6 Recoded Datasets · p. 7

    Phylogenetic analyses of early animal relationships using models that do not describe site-specific amino acid differences invariably support Ctenophora-sister, whereas analyses using models that can accommodate such preferences predominantly favor Porifera-sister.

    “Phylogenetic analyses of early animal relationships using models that do not describe site-specific amino acid differences (WAG+G, LG+G, GTR+G, and optimal PF-schemes) invariably support Ctenophora-sister, whereas analyses using models that can accommodate such preferences (e.g., CAT-GTR+G) predominantely favor Porifera-sister.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Weighing the Evidence · p. 6

    Although CAT-GTR+G describes Chang and WhelanD20 better than other models, overall model adequacy can still be improved using data recoding, and Porifera-sister was favored under all recoding strategies.

    “Our PPA tests indicate that although CAT-GTR+G describes Chang and WhelanD20 better than other models, overall model adequacy can still be improved using data recoding, and Porifera-sister was favored under all recoding strategies.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation
    § Weighing the Evidence · p. 5

    Phylogenomic analyses have supported both Porifera-sister and Ctenophora-sister.

    “However, phylogenomic analyses have supported both Porifera-sister [6, 7, 39, 41-44] and Ctenophora-sister [8-12, 35, 45, 46], with some studies (like [47]) presenting both trees.”
    Quote in the paperSubstantial claim0 of 10 sources checked
    • cites [6] Pisani, 2015, Genomic data do not support comb jellies as the sister group to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

    • cites [7] Simion, 2017, A large and consistent phylogenomic dataset supports sponges as the sister group to all other animals, Curr. Biol. source

      Not checked in the free run.

    • cites [39] Nosenko, 2013, Deep metazoan phylogeny: when different genes tell different stories, Mol. Phylogenet. Evol. source

      Not checked in the free run.

    • cites [41] Philippe, 2009, Phylogenomics revives traditional views on deep animal relationships, Curr. Biol. source

      Not checked in the free run.

    • cites [42] Pick, 2010, Improved phylogenomic taxon sampling noticeably affects nonbilaterian relationships, Mol. Biol. Evol. source

      Not checked in the free run.

    • cites [43] Philippe, 2011, Resolving difficult phylogenetic questions: why more sequences are not enough, PLoS Biol. source

      Not checked in the free run.

    • cites [44] Eitel, 2017, A taxogenomics approach uncovers a new genus in the phylum Placozoa, bioRxiv

      Not checked in the free run.

    • cites [8] Ryan, 2013, The genome of the ctenophore Mnemiopsis leidyi and its implications for cell type evolution, Science source

      Not checked in the free run.

    • cites [9] Moroz, 2014, The ctenophore genome and the evolutionary origins of neural systems, Nature source

      Not checked in the free run.

    • cites [10] Whelan, 2015, Error, signal, and the placement of Ctenophora sister to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation
    § Weighing the Evidence · p. 5

    Across-site replacement-pattern heterogeneity and compositional heterogeneity across taxa can both have a misleading effect on phylogenetic reconstruction.

    “Across-site replacement-pattern heterogeneity [26] and compositional heterogeneity across taxa [13, 55] can both have a misleading effect on phylogenetic reconstruction.”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [26] Lartillot, 2007, Suppression of long-branch attraction artefacts in the animal phylogeny using a site-heterogeneous model, BMC Evol. Biol. source

      Not checked in the free run.

    • cites [13] Hrdy, 2004, Trichomonas hydrogenosomes contain the NADH dehydrogenase module of mitochondrial complex I, Nature source

      Not checked in the free run.

    • cites [55] Foster, 2004, Modeling compositional heterogeneity, Syst. Biol. source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation
    § Weighing the Evidence · p. 5

    Ctenophora-sister invariably emerges when site-specific amino acid differences are not modeled or when they are modeled but distant outgroups are used to root the tree.

    “Ctenophora-sister invariably emerges when site-specific amino acid differences are not modeled [10] or when site-specific amino acid differences are modeled but distant outgroups are used to root the tree [6].”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [10] Whelan, 2015, Error, signal, and the placement of Ctenophora sister to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

    • cites [6] Pisani, 2015, Genomic data do not support comb jellies as the sister group to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Premisecitation
    § Weighing the Evidence · p. 5

    To minimize systematic error, it is important to use the evolutionary model that best fits the data.

    “To minimize systematic error, it is important to use the evolutionary model that best fits the data [18, 48-50].”
    Quote in the paperNot a substantial claim0 of 4 sources checked
    • cites [18] Heath, 2008, Taxon sampling and the accuracy of phylogenetic analyses, J. Syst. Evol.

      Not checked in the free run.

    • cites [48] Posada, 1998, MODELTEST: testing the model of DNA substitution, Bioinformatics source

      Not checked in the free run.

    • cites [49] Cunningham, 1998, Best-fit maximum-likelihood models for phylogenetic inference: empirical tests with known phylogenies, Evolution

      Not checked in the free run.

    • cites [50] Sullivan, 2005, Model selection in phylogenetics, Annu. Rev. Ecol. Evol. Syst. source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation
    § Testing the distribution of the signal in favor of Porifera-sister · p. 12

    Shen et al. [11] generated datasets that excluded the few genes they identified as outliers.

    “Shen et al. [11] generated datasets that excluded the few genes that they identified as ''outliers,''”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [11] Shen, 2017, Contentious relationships in phylogenomic studies can be driven by a handful of genes, Nat Ecol Evol source

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data
    § Weighing the Evidence · p. 6

    The conclusion of Shen et al. [11] that signal for Porifera-sister is limited to 'outlier genes' is rejected, and their approach is model dependent.

    “This result rejects the conclusions of [11] that signal for Porifera-sister is limited to ''outlier genes'' and indicates that their approach is model dependent: arguments about model adequacy should be considered when implementing it.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [11] Shen, 2017, Contentious relationships in phylogenomic studies can be driven by a handful of genes, Nat Ecol Evol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § Phylogenetic analyses · p. 11

    WhelanD20 was not analyzed under its optimal PF-scheme because [10] already showed that this dataset supports Ctenophora-sister when the PF-scheme is used.

    “WhelanD20 was not analyzed under its optimal PF-scheme because [10] already showed that this dataset supports Ctenophora-sister when the PF-scheme is used.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [10] Whelan, 2015, Error, signal, and the placement of Ctenophora sister to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § Testing model adequacy · p. 10

    GTR+G was used by [8].

    “GTR+G was used by [8]”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [8] Ryan, 2013, The genome of the ctenophore Mnemiopsis leidyi and its implications for cell type evolution, Science source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation
    § Testing the distribution of the signal in favor of Porifera-sister · p. 12

    Shen et al. [11] analyzed their outlier-excluded datasets under LG+G and found support for Ctenophora-sister.

    “analyzed these datasets under LG+G, and found support for Ctenophora-sister”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [11] Shen, 2017, Contentious relationships in phylogenomic studies can be driven by a handful of genes, Nat Ecol Evol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given
    § Dataset selection · p. 10

    The dataset from [10] selected for analysis had not been reassessed in [6].

    “The second is a dataset from [10] that had not been reassessed in [6].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [10] Whelan, 2015, Error, signal, and the placement of Ctenophora sister to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

    • cites [6] Pisani, 2015, Genomic data do not support comb jellies as the sister group to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § Testing model adequacy · p. 10

    WAG+G was used by [9].

    “WAG+G was used by [9]”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [9] Moroz, 2014, The ctenophore genome and the evolutionary origins of neural systems, Nature source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation
    § Weighing the Evidence · p. 6

    Branch length is the product of time and substitution rate, and even slowly evolving outgroups can be long branched if they diverged much earlier than the ingroup, potentially exacerbating long-branch attraction artifacts.

    “Branch length is the product of time and substitution rate, and even slowly evolving outgroups can be long branched if they diverged much earlier than the ingroup, potentially exacerbating long-branch attraction artifacts [18, 57].”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [18] Heath, 2008, Taxon sampling and the accuracy of phylogenetic analyses, J. Syst. Evol.

      Not checked in the free run.

    • cites [57] Hillis, 1998, Taxonomic sampling, phylogenetic accuracy, and investigator bias, Syst. Biol. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § Testing model adequacy · p. 10

    LG+G was used by [11].

    “LG+G was used by [11]”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [11] Shen, 2017, Contentious relationships in phylogenomic studies can be driven by a handful of genes, Nat Ecol Evol source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisenone given
    § Testing the effect of incrementing the number of ctenophoran lineages on the phylogenetic stability of recoded datasets · p. 12

    Whelan et al. [38] increased the number of ctenophoran species compared to those they used in [10].

    “Whelan et al. [38] have increased the number of ctenophoran species compared to those they used in [10].”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [38] Whelan, 2017, Ctenophore relationships and their placement as the sister group to all other animals, Nat. Ecol. Evol. source

      Not checked in the free run.

    • cites [10] Whelan, 2015, Error, signal, and the placement of Ctenophora sister to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductcitation
    § Phylogenetic analyses · p. 11

    Chang was analyzed under its optimal PF-scheme using RAxML version 8.2.9, with support estimated using the rapid bootstrap option (100 replicates).

    “Chang was analyzed under its optimal PF-scheme using RAxML version 8.2.9 [63], with support estimated using the rapid bootstrap option (100 replicates).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [63] Stamatakis, 2014, RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies, Bioinformatics source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given
    § Phylogenetic analyses · p. 11

    The number of samples to be discarded as burnin was independently assessed for each analysis by visually checking traces of likelihood and parameter values.

    “The number of samples to be discarded as burnin was independently assessed for each analysis by visually checking traces of likelihood and parameter values.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisereasoning
    § Testing the distribution of the signal in favor of Porifera-sister · p. 12

    Dataset D16 from [10] is very similar to WhelanD20.

    “which is very similar to WhelanD20”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [10] Whelan, 2015, Error, signal, and the placement of Ctenophora sister to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation
    § Dataset selection · p. 10

    Dataset D20 explicitly excludes proteins of significant compositional heterogeneity.

    “which apparently explicitly excludes proteins of significant compositional heterogeneity (see [10])”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [10] Whelan, 2015, Error, signal, and the placement of Ctenophora sister to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § Testing model adequacy · p. 10

    CAT-GTR+G was used by [6].

    “CAT-GTR+G was used by [6]”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [6] Pisani, 2015, Genomic data do not support comb jellies as the sister group to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning
    § Figure 1. CAT-GTR+G Trees from Dayhoff-6 Recoded Datasets · p. 7

    Ctenophora-sister is a tree reconstruction artifact, whereas a placement of the sponges as the sister group of all other animals most likely reflects genuine phylogenetic signal.

    “We conclude that Ctenophora-sister is a tree reconstruction artifact, whereas a placement of the sponges as the sister group of all other animals most likely reflects genuine phylogenetic signal.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Ctenophora-sister is currently uncorroborated by independent evidence.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § Testing model adequacy · p. 10

    PF-schemes were used by [10].

    “PFschemes were used by [10]”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [10] Whelan, 2015, Error, signal, and the placement of Ctenophora sister to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductnone given
    § Phylogenetic analyses · p. 11

    For each analysis two chains were run and convergence was assessed using the bpcomp and tracecomp tools in PhyloBayes.

    “For each analysis two chains were run and convergence was assessed using the bpcomp and tracecomp tools in PhyloBayes.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given
    § Testing model adequacy · p. 11

    All PPA analyses used a minimum of 100 replicates.

    “All PPA analyses used a minimum of 100 replicates.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § Data recoding · p. 11

    Amino acid recoding has long been considered a powerful strategy to reduce compositional heterogeneity and saturation in the data.

    “Amino acid recoding has long been considered as a powerful strategy to reduce compositional heterogeneity and saturation in the data [7, 13-15, 31-36]”
    Quote in the paperNot a substantial claim0 of 10 sources checked
    • cites [7] Simion, 2017, A large and consistent phylogenomic dataset supports sponges as the sister group to all other animals, Curr. Biol. source

      Not checked in the free run.

    • cites [13] Hrdy, 2004, Trichomonas hydrogenosomes contain the NADH dehydrogenase module of mitochondrial complex I, Nature source

      Not checked in the free run.

    • cites [14] Susko, 2007, On reduced amino acid alphabets for phylogenetic inference, Mol. Biol. Evol. source

      Not checked in the free run.

    • cites [15] Kosiol, 2004, A new criterion and method for amino acid classification, J. Theor. Biol. source

      Not checked in the free run.

    • cites [31] Rodríguez-Ezpeleta, 2007, Detecting and overcoming systematic errors in genome-scale phylogenies, Syst. Biol. source

      Not checked in the free run.

    • cites [32] Cox, 2008, The archaebacterial origin of eukaryotes, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

    • cites [33] Rota-Stabelli, 2013, Serine codon-usage bias in deep phylogenomics: pancrustacean relationships as a case study, Syst. Biol. source

      Not checked in the free run.

    • cites [34] Domman, 2015, Plastid establishment did not require a chlamydial partner, Nat. Commun. source

      Not checked in the free run.

    • cites [35] Borowiec, 2015, Extracting phylogenetic signal and accounting for bias in whole-genome data sets supports the Ctenophora as sister to remaining Metazoa, BMC Genomics source

      Not checked in the free run.

    • cites [36] Schwentner, 2017, A phylogenomic solution to the origin of insects by resolving crustacean-hexapod relationships, Curr. Biol. source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation
    § Testing the distribution of the signal in favor of Porifera-sister · p. 12

    WhelanD20 was obtained by removing genes of significant compositional heterogeneity from D16.

    “as the latter was obtained in [10] by removing genes of significant compositional heterogeneity from D16”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [10] Whelan, 2015, Error, signal, and the placement of Ctenophora sister to all other animals, Proc. Natl. Acad. Sci. USA source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationreasoning
    § A discussion of alternative evolutionary models · p. 12

    Recoding strategies can in some cases cause signal erosion, but this can be monitored because it will result in collapsed nodes that do not provide support for any topology, rather than a change in the topology supported by the data.

    “Because amino acids are grouped into classes recoding strategies can in some cases cause signal erosion. This however can be monitored because it will result in collapsed nodes that do not provide support for any topology, rather than a change in the topology supported by the data.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation
    § A discussion of alternative evolutionary models · p. 12

    Different sites and lineages accumulate substitutions at different rates.

    “Different sites and lineages accumulate substitutions at different rates (e.g., [67-69])”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [67] Yang, 1993, Maximum-likelihood estimation of phylogeny from DNA sequences when substitution rates differ over sites, Mol. Biol. Evol.

      Not checked in the free run.

    • cites [68] Thorne, 1998, Estimating the rate of evolution of the rate of molecular evolution, Mol. Biol. Evol. source

      Not checked in the free run.

    • cites [69] Yang, 1994, Maximum likelihood phylogenetic estimation from DNA sequences with variable rates over sites: approximate methods, J. Mol. Evol. source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation
    § A discussion of alternative evolutionary models · p. 12

    Different amino acids are preferentially found at different sites or in different lineages.

    “different amino acids are preferentially found at different sites or in different lineages (e.g., [26, 55])”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [26] Lartillot, 2007, Suppression of long-branch attraction artefacts in the animal phylogeny using a site-heterogeneous model, BMC Evol. Biol. source

      Not checked in the free run.

    • cites [55] Foster, 2004, Modeling compositional heterogeneity, Syst. Biol. source

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationreasoning
    § A discussion of alternative evolutionary models · p. 12

    Because Dayhoff recoding clusters amino acids with comparable properties within the same category, it reduces heterogeneity across both sites and lineages, potentially reducing the impact of systematic errors on phylogenetic estimation under currently available models.

    “Because Dayhoff recoding clusters amino acids with comparable properties within the same category, it reduces heterogeneity across both sites and lineages, potentially reducing the impact of systematic errors on phylogenetic estimation under currently available models.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § A discussion of alternative evolutionary models · p. 12

    Different proteins can have their own protein-specific replacement rates.

    “different proteins can have their own protein-specific replacement rates (e.g., [24])”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [24] Lanfear, 2012, Partitionfinder: combined selection of partitioning schemes and substitution models for phylogenetic analyses, Mol. Biol. Evol. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § Data recoding · p. 11

    Dayhoff recoding partitions amino acids into six classes or bins based on the log odds matrix of probabilities of pairs of amino acids appearing together in a PAM 250 matrix.

    “Dayhoff recoding partitions amino acids in six classes or bins (Table 2), based on the log odds matrix of probabilities of pairs of amino acids appearing together in a PAM 250 matrix”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [14] Susko, 2007, On reduced amino acid alphabets for phylogenetic inference, Mol. Biol. Evol. source

      Not checked in the free run.

    • cites [37] Dayhoff, 1978, A model of evolutionary change in proteins

      Not checked in the free run.

46 of 46

  • The paper relies on this without support.

    Ctenophora-sister is currently uncorroborated by independent evidence.

    “whereas Ctenophora-sister is currently uncorroborated by independent evidence”

  • The authors state this as an assumption.

    When computing Z-scores, the null distribution is approximately normal, with the Z-score analogous to a standard normal quantile.

    “When computing Z-scores it is assumed that the null distribution is approximately normal, with the Z-score being analogous to a standard normal quantile.”

  • The paper relies on this without support.

    Datasets without outlier genes should not include signal in favor of Porifera-sister.

    “As datasets without ''outlier genes'' should not include signal in favor of Porifera-sister”

  • The paper relies on this without support.

    Current implementations of models that explicitly take into consideration across-taxon compositional heterogeneity are computationally too demanding to allow their application to large phylogenomic datasets.

    “However, current implementations of these models are computationally too demanding to allow their application to large phylogenomic datasets.”
4 of 4

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • The RL2 findings support recent efforts pairing recoding and site-heterogeneous models to resolve difficult, ancient nodes in animal phylogeny, despite the erosion of data relevant to more shallow nodes.

    “The RL2 findings also support recent efforts pairing recoding and site-heterogeneous models to resolve difficult, ancient nodes in animal phylogeny5,9,24,33, despite the erosion of data relevant to more shallow nodes.”

    From Evidence for sponges as sister to all other animals from partitioned phylogenomics with mixture models and recoding (2021) · cites this paper as [9]

    In Weighing the Evidence (p. 6), the paper reports that Dayhoff-6 recoding paired with the site-heterogeneous CAT-GTR+G model strongly supported Porifera as sister to other animals, an ancient node. A discussion of alternative evolutionary models (p. 12) says recoding can cause signal erosion and collapsed nodes, but does not identify shallower nodes as particularly affected. (LLM, conf=0.83)

    Experimental — accuracy not yet measured

  • Saturation and convergence are often underestimated when using standard site-homogeneous models, causing LBA.

    “By contrast, saturation and convergence are often underestimated when using standard site-homogeneous models, causing LBA8,9,22.”

    From Evidence for sponges as sister to all other animals from partitioned phylogenomics with mixture models and recoding (2021) · cites this paper as [9]

    The Summary (p. 4) finds that the tested models inadequately describe site-specific amino-acid preferences, including a measure of convergence; Weighing the Evidence (p. 6) links poorer model adequacy and long branches to potentially artifactual trees. The paper does not establish that standard site-homogeneous models specifically underestimate both saturation and convergence, or that this underestimation causes LBA. (LLM, conf=0.84)

    Experimental — accuracy not yet measured

  • Most current analyses find ctenophores to be the sister group either of all other metazoans (Ctenophora-sister hypothesis) or of all animals except sponges (Porifera-sister hypothesis).

    “most current analyses instead find ctenophores to be the sister group either of all other metazoans (Ctenophora-sister hypothesis) or of all animals except sponges (Porifera-sister hypothesis) [4–6, 14”

    From Cambrian Sessile, Suspension Feeding Stem-Group Ctenophores and Evolution of the Comb Jelly Body Plan (2019) · cites this paper as [5]

    The Summary (p. 3) and “Weighing the Evidence” (p. 5) support the Ctenophora-sister part: they describe ctenophores as sister to all other animals and report analyses supporting that hypothesis. But they define Porifera-sister as sponges being sister to all other animals, not as ctenophores being sister to all animals except sponges. (LLM, conf=0.89)

    Experimental — accuracy not yet measured

  • CAT/CATGTR offers improved resilience to long-branch attraction artefacts and is almost always better fitting than standard site-homogeneous amino acid substitution models.

    “CAT/CATGTR offers improved resilience to long-branch attraction artefacts (LBA)22 and is almost always better fitting than standard site-homogeneous amino acid substitution models (e.g. WAG)8–10,23–29.”

    From Evidence for sponges as sister to all other animals from partitioned phylogenomics with mixture models and recoding (2021) · cites this paper as [9]

  • The branch leading to comb jellies is extremely long, and when re-analyses of the same datasets are performed using methods intended to alleviate systematic errors, such as site-heterogeneous models and/or recoding, sponges tend to be recovered as sister to other animals.

    “However, the branch leading to comb jellies is extremely long, and when re-analyses of the same datasets are performed using methods intended to alleviate systematic errors, such as site-heterogeneous models (i.e. CAT/CATGTR) and/or recoding, sponges tend to be recovered as sister to other animals (Fig. 1a)8–10.”

    From Evidence for sponges as sister to all other animals from partitioned phylogenomics with mixture models and recoding (2021) · cites this paper as [9]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Some recoded Chang analyses have higher PPA-MAX Z scores

    Minor

    The Chang GTR+G PPA-MAX Z score is 18.84 without recoding, but rises to 26.99 with Dayhoff-6 recoding and 28.48 with S&R-6 recoding. Those results contradict the claim that all recoded datasets had lower PPA-MAX Z scores.

  • WhelanD20 alignment length differs by 4,000 positions

    Minor

    Table 1 describes WhelanD20-Opistho as having 46,542 amino acid positions, while the methods say WhelanD20 has 42,542 positions. WhelanD20-Opistho is identified as the version retaining all original outgroups, not a different-length alignment.

2 of 2

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Feuda, R., Dohrmann, M., Pett, W., Philippe, H., Rota-Stabelli, O., Lartillot, N., Wörheide, G., & Pisani, D.
Affiliations
  • California Institute of Technology (US)
  • Ludwig-Maximilians-Universität München (DE)
  • Iowa State University (US)
  • Centre National de la Recherche Scientifique (FR)
  • Lyon 1 Université (FR)
  • Bayerische Staatssammlung für Paläontologie und Geologie (DE)
  • University of Bristol (GB)
Corresponding author
  • Wörheide, Gert
  • Pisani, Davide
ORCID
Venue
Current Biology
Publisher
Elsevier BV
Published
2017-12
Volume
27
Issue
24
Pages
3864-3870.e4
ISSN
0960-9822 (ISSN-L)
DOI
10.1016/j.cub.2017.11.008
OpenAlex
W2775427740
Semantic Scholar
c5ee9d8995148b11f400719b18bbe60028a76e97
Cited by
291
References
72
Citations by year
2026: 10 · 2025: 22 · 2024: 25 · 2023: 23 · 2022: 29 · 2021: 49 · 2020: 37 · 2019: 56 · 2018: 40
FWCI
25.85
Type
journal article
Language
en
Open access
Licence
Locations
  • Current Biology (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Explore Bristol Research (publishedVersion, cc-by)
  • CINECA IRIS Institutional Research Information System (Fondazione Edmund Mach) (submittedVersion, cc-by)
  • CaltechAUTHORS (California Institute of Technology) (submittedVersion, cc-by)
  • HAL (submittedVersion)
  • Archivio istituzionale della ricerca (Alma Mater Studiorum Università di Bologna) (submittedVersion)
  • Figshare (submittedVersion, cc-by)
  • Iris (University of Trento) (submittedVersion, cc-by)
  • Bristol Research (University of Bristol) (submittedVersion, cc-by)
Funders
  • NERC (NE/P013643/1)
  • Templeton Foundation (60579)
  • French Laboratory of Excellence (ANR-10-LABX- 41, ANR-11-IDEX-0002-02)
  • French National Research Agency (ANR-10-BINF-01-01)
  • LMU Munich’s Institutional Strategy LMUexcellent
  • German Research Foundation (DFG) (Wo896/15-1)
Topics
  • Physical Sciences › Earth and Planetary Sciences › Paleontology › Marine Invertebrate Physiology and Ecology
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Phylogenetic Studies
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protist diversity and phylogeny
Keywords
animal phylogeny, sponges, comb jellies, Ctenophora, sister group relationships, compositional heterogeneity, data recoding, topology change
MeSH terms
Animals, Biological Evolution, Phylogeny, Porifera, Sequence Analysis, Protein

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-06.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-06.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review