Stability, sensitivity, science and heurism

We examine recently proposed justifications of sensitivity analysis sensu Wheeler (1995 ), here referred to as weighted‐classes sensitivity analysis (WCSA). We refute Giribet's (2003a) claim that WCSA is the strictest possible test for a given phylogenetic hypothesis. Giribet's (2003a) classification of data exploration methods as evaluating “nodal stability” or “nodal support” is arbitrary, at odds with common usage and actually obscures the relationships between the methods he examined, all of which seek to assess the sensitivity of results to variation in analytical conditions. Stability, whether statistical or taxonomic, is not a goal of phylogenetic science. Statistical stability necessarily involves trial repetition, which is impossible in ideographic sciences like phylogenetics. Taxonomic stability can be nothing more than an unintended by‐product of scientific inquiry, i.e., the repeated failure to refute a hypothesis. Schulmeister's (2003 ) “robust‐choice” defense of WCSA does not succeed in placing non‐arbitrary bounds on parameters, and her interpretation of this approach as simultaneously verificationist and falsificationist is logically inconsistent. WCSA is neither scientific nor heuristic and therefore does not contribute to the advancement of objective knowledge claims. © The Willi Hennig Society 2005.     

Phylogenetic analysis of Themira (Sepsidae: Diptera): sensitivity analysis, alignment, and indel treatment in a multigene study

In this study we use sensitivity analysis sensu Wheeler (1995 ) for a matrix entirely composed of DNA sequences. We propose that not only congruence but also phylogenetic structure, as measured by character resampling, should be used to choose among competing weighting regimes. An extensive analysis of a five‐gene data set for Themira (Sepsidae: Diptera) reveals that even with different ways of partitioning the data, measures of topological congruence, character incongruence, and phylogenetic structure favor similar weighting regimes involving the down‐weighting of transitions. We furthermore use sensitivity analysis for obtaining empirical evidence that allows us to select weights for third positions, deciding between treating indels as fifth character states or missing values, and choosing between manual and computational alignments. For our data, sensitivity analysis favors manual alignment over a Clustal‐generated numerical alignment, the treatment of indels as fifth character states over considering them missing values, and equal weights for all positions in protein‐encoding genes over the down‐weighting of third positions. Among the topological congruence measures compared, symmetric tree distance performed best. Partitioned Bremer Support analysis reveals that COI contributes the largest amount of support for our phylogenetic tree for Themira. © The Willi Hennig Society 2005.     

On gaps.

Gaps result from the alignment of sequences of unequal length during primary homology assessment. Viewed as character states originating from particular biological events (mutation), gaps contain historical information suitable for phylogenetic analysis. The effect of gaps as a source of phylogenetic data is explored via sensitivity analysis and character congruence among different data partitions. Example data sets are provided to show that gaps contain important phylogenetic information not recovered by those methods that omit gaps in their calculations. However, gap cost schemes are arbitrary (although they must be explicit) and thus data exploration is a necessity of molecular analyses, while character congruence is necessary as an external criterion for hypothesis decision.     

Sequence length variation, indel costs, and congruence in sensitivity analysis

The behavior of two topological and four character‐based congruence measures was explored using different indel treatments in three empirical data sets, each with different alignment difficulties. The analyses were done using direct optimization within a sensitivity analysis framework in which the cost of indels was varied. Indels were treated either as a fifth character state, or strings of contiguous gaps were considered single events by using linear affine gap cost. Congruence consistently improved when indels were treated as single events, but no congruence measure appeared as the obviously preferable one. However, when combining enough data, all congruence measures clearly tended to select the same alignment cost set as the optimal one. Disagreement among congruence measures was mostly caused by a dominant fragment or a data partition that included all or most of the length variation in the data set. Dominance was easily detected, as the character‐based congruence measures approached their optimal value when indel costs were incremented. Dominance of a fragment or data partition was overwhelmed when new sequence length‐variable fragments or data partitions were added. © The Willi Hennig Society 2005.     

Sensitivity analysis, molecular systematics and natural history evolution of Scathophagidae (Diptera: Cyclorrhapha: Calyptratae)

The 60 000 described species of Cyclorrhapha are characterized by an unusual diversity in larval life‐history traits, which range from saprophagy over phytophagy to parasitism and predation. However, the direction of evolutionary change between the different modes remains unclear. Here, we use the Scathophagidae (Diptera) for reconstructing the direction of change in this relatively small family (≈ 250 spp.) whose larval habits mirror the diversity in natural history found in Cyclorrhapha. We subjected a molecular data set for 63 species (22 genera) and DNA sequences from seven genes (12S, 16S, Cytb, COI, 28S, Ef1‐alfa, Pol II) to an extensive sensitivity analysis and compare the performance of three different alignment strategies (manual, Clustal, POY). We find that the default Clustal alignment performs worst as judged by character incongruence, topological congruence and branch support. For this alignment, scoring indels as a fifth character state worsens character incongruence and topological congruence. However, manual alignment and direct optimization perform similarly well and yield near‐identical trees, although branch support is lower for the direct‐optimization trees. All three alignment techniques favor the upweighting of transversion. We furthermore confirm the independence of the concepts “node support” and “node stability” by documenting several cases of poorly supported nodes being very stable and cases of well supported nodes being unstable. We confirm the monophyly of the Scathophagidae, its two constituent subfamilies, and most genera. We demonstrate that phytophagy in the form of leaf mining is the ancestral larval feeding habit for Scathophagidae. From phytophagy, two shifts to saprophagy and one shift to predation has occurred while a second origin of predation is from a saprophagous ancestor. © The Willi Hennig Society 2006.     

War and peace in phylogenetics: a rejoinder on total evidence and consensus

For more than 10 years, systematists have been debating the superiority of character or taxonomic congruence in phylogenetic analysis. In this paper, we demonstrate that the competing approaches can converge to the same solution when a consensus method that accounts for branch lengths is selected. Thus, we propose to use both methods in combination, as a way to corroborate the results of combined and separate analyses. This so-called "global congruence" approach is tested with a wide variety of examples sampled from the literature, and the results are compared with those obtained by standard consensus methods. Our analyses show that when the total evidence and consensus trees differ topologically, collapsing weakly supported nodes with low bootstrap support usually improves "global congruence".     

Congruence,non‐homology,and the phylogeny of basal turtles

Joyce, W.G. and Sterli J. 2010. Congruence, non‐homology, and the phylogeny of basal turtles.–Acta Zoologica (Stockholm) Modern cladistic analysis is characterized by the assembly of increasingly larger data sets coupled with the use of congruence as the final test of homology. Some critics of this development have recently called for a return to more detailed primary homology analysis while questioning the utility of congruence. This discussion appears to be central to the debate regarding the phylogenetic relationships of basal turtles, as the large data sets developed by us have been criticized recently for utilizing poorly constructed characters and including too many homoplasy‐prone characters. Our analysis of this critique reveals that (1) new information regarding poorly understood taxa has a greater impact on the outcome of turtle phylogenies than the characters under dispute; (2) most current turtle phylogenies differ in taxon sampling, not character sampling, and so it appears illogical to condemn a particular analysis for its character sampling; (3) even evolutionary taxonomists should agree that key characters utilized to resolve basal turtle relationships cannot be thought to be ‘infallible’; (4) whereas various criteria provide positive evidence for homology, only congruence provides positive evidence for non‐homology; and (5) a stalemate between conflicting camps within a congruence frame work is preferable to the ad hoc dismissal of data sets, because authoritative statements are untestable.     

Size, frequency, and phylogenetic signal of multiple-residue indels in sequence alignment of introns

Indels in DNA sequences frequently affect more than a single nucleotide, creating problems for alignment, character coding and phylogenetic analysis. However, the size and frequency of multiple‐residue indels is not usually tested, and with popular alignment packages their reconstruction is indirectly acheived by reducing the affine (gap extension) cost. We explored the length distribution of indels in intron sequences of the gene Mp20 by modifying the gap opening and gap extension costs. Given a “known” tree for the study group, global homology levels were greatest under low gap cost, with gap extension costs of roughly 0.4‐fold the opening cost. Different approaches to gap coding and weighting suggested that taxonomic congruence was correlated with high frequencies of multiple‐position indels, with a maximum indel length of 2–5 bp and few indels above 15 bp, but also including a proportion of indels > 100 bp. Only a small minority of indels could be reconstructed as single‐position indels. Consequently, tree topologies improved when homologous multinucleotide indels were recoded as binary characters which are otherwise highly homoplastic and weighted characters in single‐position coding. In tree‐generating alignment procedures as implemented in POY, where gap penalty determines the character weight during tree search, the problem of assigning inappropriately high weight to multiple‐residue indels could partly be overcome by setting the extension costs to about 0.4‐fold lower than gap opening costs. We conclude that multiple consecutive gap positions are not independent characters and hence methods for parsimony reconstruction of long indels are required. Finally, we also observed a general lack of correlation between taxonomic and character congruence, demonstrating the difficulties of applying congruence criteria to decide among competing alignments. This highlights the value of recent model‐based alignment procedures which can implement the statistical distributions of indel size classes, and do not rely on potentially circular strategies for optimizing overall congruence. © The Willi Hennig Society 2006.     

Soft anatomy, diffuse homoplasy, and the relationships of lizards and snakes

Most previous phylogenetic analyses of squamates (‘lizards’ and snakes) employing large character sets have focused on osteology. Soft anatomical traits bearing on this problem have usually been considered in small subsets. Here, a comprehensive phylogenetic analysis of squamate soft anatomy is attempted. 126 informative characters are assessed for 23 squamate lineages, representing snakes, amphisbaenians, dibamids, and all the traditionally recognized ‘families’ of lizards. The traditionally recognized groupings Iguania, Scleroglossa, Gekkota, Scincomorpha, Anguimorpha and Varanoidea are corroborated in this analysis. More controversial taxa are resolved as follows. Xantusiids, amphisbaenians and dibamids cluster with gekkotans, and snakes are strongly allied with anguimorphs in general, and varanids in particular. Nearly all these clades are congruent with those found in a recent comprehensive osteological analysis; the strong support for snake‐varanid relationships found in both studies is particularly notable. This congruence is surprising given that previous studies of soft anatomy tended to give differing and often heterodox results. These previous results can be attributed to overrepresentation of misleading characters in small isolated data sets. Such misleading signals are minimized when data sets are combined. For instance, the snake‐varanid clade is contradicted by many characters, and analyses of particular organ systems therefore give differing results. However, characters that are incongruent with the snake‐varanid clade also disagree with each other (diffuse homoplasy), rather than forming coherent support for some particular alternative clade (concerted homoplasy). In a combined analysis these incongruent but diffuse characters cancel each other out to leave a very strong (and orthodox) phylogenetic signal. These results underscore the view that the raw amount of homoplasy — as revealed by consistency and retention indices — is not the only determinant of phylogenetic signal; the distribution of that homoplasy is also important. Thus, questioning a phylogenetic hypothesis (e.g. the snake‐varanid clade) by identifying numerous conflicting characters is insufficient — the structure of the conflicting characters should be assessed in a rigorous phylogenetic analysis.     

Corroboration among Data Sets in Simultaneous Analysis: Hidden Support for Phylogenetic Relationships among Higher Level Artiodactyl Taxa

In the taxonomic congruence approach to systematics, data sets are analyzed separately, and corroboration among data sets is indicated by replicated components in topologies derived from the separate analyses. By contrast, in the total evidence and conditional combination approaches, characters from different data sets are mixed in combined phylogenetic analyses. In optimal topologies derived from such simultaneous analyses, support for a particular node may be attributed to one, some, or all of the individual data sets. Partitioned branch support (PBS) is one technique for describing the distribution of character support and conflict among data sets in simultaneous analysis. PBS is analogous to branch support (BS), but recognizes hidden support and conflicts that emerge with the combination of characters from different data sets. For both BS and PBS, support for a particular node is interpreted as the difference in cost between optimal and suboptimal topologies. A different measure, the clade stability index (CSI), assesses the robustness of a particular node through the successive removal of characters. Here, we introduce variations of the CSI, the data set removal index (DRI) and nodal data set influence (NDI), that indicate the stability of a particular node to the removal of entire data sets. Like PBS, the DRI and NDI summarize the influence of different data sets in simultaneous analysis. However, because these new methods and PBS use different perturbations to assess stability, DRI and NDI scores do not always predict PBS scores and vice versa. In this report, the DRI and NDI are compared to PBS and taxonomic congruence in a cladistic analysis of 17 data sets for Artiodactyla (Mammalia). Five indices of hidden support and conflict are defined and applied to the combined artiodactyl character set. These measures identify substantial hidden support for controversial relationships within Artiodactyla. Hidden character support is ignored in the taxonomic congruence approach to systematics, but the DRI, NDI, and PBS utilize this cryptic information in estimates of support among data sets for a given node.     

Problematic character coding methods in morphology and their effects

The effects of different coding practices in morphological phylogenetic analysis are well documented. In many cases, we can determine that certain practices can be regarded as undesirable and should be avoided. Certain coding practices do not correctly translate the expected information to the cladistic algorithm. It may go unnoticed that expressions of character information in character lists, which may be entirely logical to any reader, do not necessarily reflect the mathematics employed by a phylogenetic algorithm. Despite a wealth of literature on coding procedures and documentation of these issues, problematic character coding practices are still common. A review is provided of different coding and character formulation practices, particularly relating to multistate character information that may either: (1) lead to a failure to capture grouping information implied in the character list; (2) cause problematic weighting or spuriously high certainty in particular optimizations; and (3) impose congruence artificially, by linking more than one variable character to a particular state. Each of these is reviewed and presented with a hypothetical example. Recommendations for avoiding these pitfalls are described in light of how parsimony algorithms work with character data. Character lists must be drawn up not only to present character variation logically, but also with consideration for how computer algorithms implement cladistic logic. The widespread use of problematic character coding procedures may account for some of the perceived problems with morphological data. Therefore, an exploration of the effects of these methods and standardization of methods should be a goal for the very near future. © 2011 The Linnean Society of London, Biological Journal of the Linnean Society, 2011, 104 , 489–498.     

Ecdysozoa versus Articulata: clades, artifacts, prejudices

The claim that monophyly of the Ecdysozoa is caused by chance similarities in 18S rDNA sequences ( Wägele et al., J. Zool. Syst. Evol. Res. 37, 211–223, 1999 ) is re-analysed from the cladistic point of view. It is shown that the molecular characters supporting the Ecdysozoa do not behave as 'noisy' in empirical studies that use the sensitivity analysis and character congruence approaches. The 'anti-noise' methodology proposed by Wägele et al. (1999) is unable to identify true misinformative data. The monophyly of the Articulata (= Annelida + Panarthropoda), proposed by Wägele et al. (1999) , is contradicted by all molecular data that support either Ecdysozoa (including Panarthropoda), or Lophotrochozoa (including Annelida), or usually both.     

The relative sensitivity of different alignment methods and character codings in sensitivity analysis

Sensitivity analysis provides a way to measure robustness of clades in sequence‐based phylogenetic analyses to variation in alignment parameters rather than measuring their branch support. We compared three different approaches to multiple sequence alignment in the context of sensitivity analysis: progressive pairwise alignment, as implemented in MUSCLE; simultaneous multiple alignment of sequence fragments, as implemented in DCA; and direct optimization followed by generation of the implied alignment(s), as implemented in POY. We set out to determine the relative sensitivity of these three alignment methods using rDNA sequences and randomly generated sequences. A total of 36 parameter sets were used to create the alignments, varying the transition, transversion, and gap costs. Tree searches were performed using four different character‐coding and weighting approaches: the cost function used for alignment or equally weighted parsimony with gap positions treated as missing data, separate characters, or as fifth states. POY was found to be as sensitive, or more sensitive, to variation in alignment parameters than DCA and MUSCLE for the three empirical datasets, and POY was found to be more sensitive than MUSCLE, which in turn was found to be as sensitive, or more sensitive, than DCA when applied to the randomly generated sequences when sensitivity was measured using the averaged jackknife values. When significant differences in relative sensitivity were found between the different ways of weighting character‐state changes, equally weighted parsimony, for all three ways of treating gapped positions, was less sensitive than applying the same cost function used in alignment for phylogenetic analysis. When branch support is incorporated into the sensitivity criterion, our results favour the use of simultaneous alignment and progressive pairwise alignment using the similarity criterion over direct optimization followed by using the implied alignment(s) to calculate branch support.     

Optimal taxonomic groups for biodiversity assessment: a meta‐analytic approach

A fundamental decision in biodiversity assessment is the selection of one or more study taxa, a choice that is often made using qualitative criteria such as historical precedent, ease of detection, or available technical or taxonomic expertise. A more robust approach would involve selecting taxa based on the a priori expectation that they will provide the best possible information on unmeasured groups, but data to inform such hypotheses are often lacking. Using a global meta‐analysis, we quantified the proportion of variability that each of 12 taxonomic groups (at the Order level or above) explained in the richness or composition of other taxa. We then applied optimization to matrices of pairwise congruency to identify the best set of complementary surrogate groups. We found that no single taxon was an optimal surrogate for both the richness and composition of unmeasured taxa if we used simple methods to aggregate congruence data between studies. In contrast, statistical methods that accounted for well‐known drivers of cross‐taxon congruence (spatial extent, grain size, and latitude) lead to the prioritization of similar surrogates for both species richness and composition. Advanced statistical methods were also more effective at describing known ecological relationships between taxa than simple methods, and show that congruence is typically highest between taxonomically and functionally dissimilar taxa. Birds and vascular plants were most frequently selected by our algorithm as surrogates for other taxonomic groups, but the extent to which any one taxon was the ‘optimal’ choice of surrogate for other biodiversity was highly context‐dependent. In the absence of other information – such as in data‐poor areas of the globe, and under limited budgets for monitoring or assessment – ecologists can use our results to assess which taxa are most likely to reflect the distribution of the richness or composition of ‘total’ biodiversity.     

Combining molecular and morphological analyses of water strider phylogeny (Hemiptera–Heteroptera, Gerromorpha): effects of alignment and taxon sampling

Abstract. The semiaquatic bugs (Hemiptera–Heteroptera, infraorder Gerromorpha), comprising water striders and their allies (c. 1900 described species), are familiar inhabitants of water surfaces in all continents. Recent fossil evidence indicates that the evolutionary history of semiaquatic bugs spans more than 120 million years of geological time. At present, our insight into the phylogeny of higher taxa is based upon Andersen's manual cladistic analysis of a suite of morphological characters. The present work expands the phylogenetic insight with numerical cladistic analyses of morphological and molecular datasets (partial sequences of 16S and 28S rDNA) for forty species of Gerromorpha covering most higher taxa (families, subfamilies), estimates of branch support, character incongruence, and topological congruence (nodal stability). For the molecular data we apply different alignment options (manual vs numerical alignment; multiple alignment vs direct optimization), treat insertion–deletion events (indels) as either missing data or as a fifth character state, subject the data to a sensitivity analysis, and estimate topological congruence between different analysis trees. Relationships change considerably under different analysis conditions, which means that there is little node stability, and for selecting preferred analysis conditions there is conflicting evidence from rescaled incongruence length difference and the key node criterion. Based on the analysis of the combined morphological and molecular datasets, this study supports the close relationship between the families Gerridae, Hermatobatidae and Veliidae (superfamily Gerroidea), but not the monophyly of the family Veliidae. The results suggest that the genus Ocellovelia (Ocelloveliinae) should be excluded from this family and placed as a sister group to Gerridae + the remaining species of Veliidae. Our study also supports a close relationship between the subfamilies Halobatinae and Ptilomerinae (Gerridae), and that the subfamily Veliinae is probably nonmonophyletic.     

Total evidence, consensus, and bat phylogeny: A distance-based approach

Resolution of the total evidence (i.e., character congruence) versus consensus (i.e., taxonomic congruence) debate has been impeded by (1) a failure to employ validation methods consistently across both tree-building and consensus analyses, (2) the incomparability of methods for constructing as opposed to those for combining trees, and (3) indifference to aspects of trees other than their topologies. We demonstrate a uniform, distance-based approach which allows for comparability among the results of character- and taxonomic-congruence studies, whether or not an identical suite of taxa has been included in all contributing data sets. Our results indicate that total-evidence and consensus trees differ little in topology if branch lengths are taken into account when combining two or more trees. In addition, when character-state data are converted to distances, our method permits their combination with information produced by techniques which generate distances directly. Moreover, treating all data sets or trees as distance matrices avoids the problem that different numbers of characters in contributing studies may confound the conclusions of a total-evidence or consensus analysis. Our protocol is illustrated with an example involving bats, in which the three component studies based on serology, DNA hybridization, and anatomy imply distinct phylogenies. However, the total-evidence and consensus trees support a fourth, somewhat different, topology resolved at all but one node and which conforms closely to the currently accepted higher category classification of Chiroptera.     

Exploring the Behavior of POY, a Program for Direct Optimization of Molecular Data

The performance of the computer program for phyloge netic analysis, POY, and its two implemented methods, "optimization alignment" and "fixed-states optimization," are explored for four data sets. Four gap costs are analyzed for every partition; some of the partitions (the 18S rRNA) are treated as a single fragment or in increasing fragments of 3, 10, and 30. Comparisons within and among methods are undertaken according to gap cost, number of fragments in which the sequences are divided, tree length, character congruence, topological congruence, primary homology statements, and computation time.     

Population genomic evidence for multiple Pliocene refugia in a montane‐restricted harvestman (Arachnida,Opiliones, Sclerobunus robustus) from the southwestern United States

The integration of ecological niche modelling into phylogeographic analyses has allowed for the identification and testing of potential refugia under a hypothesis‐based framework, where the expected patterns of higher genetic diversity in refugial populations and evidence of range expansion of nonrefugial populations are corroborated with empirical data. In this study, we focus on a montane‐restricted cryophilic harvestman, Sclerobunus robustus, distributed throughout the heterogeneous Southern Rocky Mountains and Intermontane Plateau of southwestern North America. We identified hypothetical refugia using ecological niche models (ENMs) across three time periods, corroborated these refugia with population genetic methods using double‐digest RAD‐seq data and conducted population‐level phylogenetic and divergence dating analyses. ENMs identify two large temporally persistent regions in the mid‐latitude highlands. Genetic patterns support these two hypothesized refugia with higher genetic diversity within refugial populations and evidence for range expansion in populations found outside hypothesized refugia. Phylogenetic analyses identify five to six genetically divergent, geographically cohesive clades of S. robustus. Divergence dating analyses suggest that these separate refugia date to the Pliocene and that divergence between clades pre‐dates the late Pleistocene glacial cycles, while diversification within clades was likely driven by these cycles. Population genetic analyses reveal effects of both isolation by distance (IBD) and isolation by environment (IBE), with IBD more important in the continuous mountainous portion of the distribution, while IBE was stronger in the populations inhabiting the isolated sky islands of the south. Using model‐based coalescent approaches, we find support for postdivergence migration between clades from separate refugia.     

Paleobotany in cladistics and cladistics in paleobotany: enlightenment and uncertainty

Data on fossil taxa can, and should, be incorporated into cladistic analyses. Potential problems with such analyses include large amounts of missing data, and uncertainty about homology of parts that are present. Ambiguity of character data may also occur with extant taxa, but rarely to the extent that it occurs in fossil data. Such ambiguity reduces the strength of the test of character congruence among taxa, in effect relaxing the criterion of parsimony. In order to minimize such effects, composite fossil taxa should be avoided when possible, and polymorphisms reduced by breaking terminals into monomorphic subunits. When results including fossils differ radically from those that exclude fossils, such differences should be approached with caution, keeping in mind the reduced strength of the parsimony analysis when large numbers of cells in a matrix are scored as ambiguous. At this point, there is no simple way to compare the “strength” of parsimony between two data sets that have different numbers of characters and/or taxa in relation to missing data. However, methods under development may provide ways to incorporate the effect of missing values into relative measures of group support such as Bremer support, character removal, and the bootstrap.