Computational protocol: Invasions but not extinctions change phylogenetic diversity of angiosperm assemblage on southeastern Pacific Oceanic islands

Similar protocols

Protocol publication

[…] Based on the checklist of species recorded for the islands, we assembled a phylogenetic tree with 921 angiosperm taxa using PHYLOMATIC [] ( accessed January 12, 2017) using tree version R20120829 is based on the APG III phylogenetic classification of angiosperms []. Within families, the phylogenetic relationships among genera were resolved through reordering by hand the obtained topology based on published phylogenies (the “graft” method, []) using Mesquite 2.74 (available at URL The full list of consulted references is provided in . The resulting topology was age-calibrated based on the divergence times of angiosperms reported by [] using the BLADJ algorithm implemented in PHYLOCOM []. We obtained a tree with 921 tips and 654 internal nodes, which accounts for 71% of the internal nodes resolved (). Complementarily, we assembled a tree exclusive to native plants (205 taxa) following the same methodology; in this tree, 74% of the internal nodes were resolved (). Although measures of community phylogenetic diversity are more sensitive to loss of resolution basally in the phylogeny, and less sensitive to loss of resolution terminally [], our levels of unresolved nodes was relatively low and should not significantly impact estimated indexes. On the other hand, the use of PHYLOMATIC and the graft method is controversial (see [] for an analytical criticism of this method).We chose this tool because several of our species are extinct or do not have genetic sequences available in GenBank. [...] Phylo-α-diversity was estimated for each island according to pre-European and current flora, based on a tree branch length index (Faith’s index []) and two distance-based indexes: the mean phylogenetic distance (MPD) and the mean nearest taxon distance (MNTD) []. Faith’s index is the sum of the branch lengths connecting all species in an assemblage []. The MPD is the mean of all phylogenetic distances that occur between all species pairs within an assemblage [,] and is interpreted as an indicator of the ‘deeper’ phylogenetic diversity of the studied flora [,]. The MNND is the mean of the phylogenetic distances that each terminal node has with their closest relatives within an assemblage [,,]. It is interpreted as the ‘terminal’ phylogenetic diversity, which reflects the influence of relatively ‘recent’ events (e.g., local extinctions, migrations) on assemblage structure []. All these indexes were independently estimated for floras of the pre-European time (original flora, including extinct species) and current floras (actual flora composition, excluding extinct species and including exotic plants). In addition, indexes were estimated exclusively to native plant species for each studied period, to compare the contribution of extinctions on phylo-diversity. Indexes were estimated based on a calibrated tree (as a chronogram) and then expressed in millions of years. Because all island assemblages showed different levels of species richness (), indexes were standardized, which allowed removal of biases associated with differences in species richness []. Standardized effect size (SES) was estimated as (observed–meannull)·(sdnull)-1, where observed is the directly estimated index, meannull is the mean of the indexes obtained after 999 randomizations and sdnull is the standard deviation of the estimated indexes after randomization. Standardized MPD (MPDSES) and MNTD (MNTDSES) also as known as net relatedness index (NRI) and nearest taxon index (NTI), respectively [, ]. These indexes are estimators of phylogenetic clustering (NRI or NTI > 1) or overdispersion (NRI or NTI < 1) [, ]. We assessed whether 95% confidence intervals of NRI and NTI obtained from pooled island (N = 6 islands) were greater than 1.96 (phylogenetic clustering) or lower than -1.96 (phylogenetic overdispersion). For all standardizations, we randomized the phylogeny tip labels across all taxa included in the trees, we used this approximation to maintain the spatial structure of species in the system. Other constrained null models that focus on randomizing the community data matrix, rather than the phylogeny, generate that any spatial contagion or dispersal limitation is not maintained in the null community data matrices []. For all our estimations, we used the picante package [] to R 3.0 (R Development Core Team).We performed two kinds of analyses with the obtained indexes as well as plant richness. For the non-standardized indexes, we estimated the percent of change for each island by comparing the pre-European and current floras; we then assessed whether these percentages deviated from the null expectation (μ = 0) using a two-tailed t-test. For the standardized index (SES), we performed a paired t-test that contrasted the obtained values between the pre-European and current floras. These analyses were performed to compare pre-European and current times for complete floras and to only native species. […]

Pipeline specifications

Software tools Phylomatic, Mesquite, Phylocom, Picante
Application Phylogenetics