Computational protocol: Characterizing the Phylogenetic Tree Community Structure of a Protected Tropical Rain Forest Area in Cameroon

Similar protocols

Protocol publication

[…] A phylogenetic tree of 372 species was generated in three steps. First, all species were grafted onto a comprehensive phylogenetic tree using the program PHYLOMATIC v3 ( The program generated a tree in which the family relationships of the sampled species followed the angiosperm phylogeny APG III version R20120829. We then manually resolved the generic relationships within most of the families based on specific molecular phylogenies (), using the software Mesquite . Only relationships that were supported with bootstrap values of more than 70% were taken into account. For families where no phylogenetic information was available or for which the published phylogeny did not provide enough insights into the relationships between genera, generic relationships where left unresolved (polytomies). Finally, we used the branch adjustment algorithm BLADJ implemented in Phylocom to scale the branch lengths based on a set of node age estimates from several publications (). For this part, we first used the dated phylogeny of Wikstrom et al for major nodes. We also used family level dated trees to further constrain certain nodes (). Intra- and interspecific branch lengths were assumed to be 0 (i.e. relationships between species and within species are unknown and unresolved). [...] To test for species turnover and phylogenetic structure, we used 3 models of randomization (, ). In model 1-3x, we randomized individuals among transects or species within each habitat type (). In model 2–3x, we randomized individuals or species among habitats (). These models of permutation aimed to test for taxonomic turnover using IST. The models 3x in which community composition was randomized but not the position of taxa in the phylogeny has been shown to be biased to test the phylogenetic structure . Thus, to test for phylogenetic structure, we used a third permutation model (model 1s, b, d, e). The model 1s randomizes the observed species across the tips of the phylogenetic tree and allowed testing for phylogenetic structure using BST and ΠST ().We undertook 999 permutations for each model, providing 999 estimations of the above differentiation coefficients under those null models. Deviations of observed coefficient from random coefficients were used to test whether IST = 0, BST = 0 or ΠST  = 0. A significant test for IST is expected at least under hypothesis (i) between transects or habitats with the model 1–3x. Under hypothesis (ii) (habitat filtering dominates), we expect BST>0 and ΠST >0 with the whole dataset- habitat (model 1s); under hypothesis (iii) (competitive exclusion between related species dominates), we expect BST<0 and ΠST <0 between transects at least within TPF habitat; while no phylogenetic significant tests should be obtained under hypotheses (i, neutral assembly with limited dispersal) and (iv, compensation between ii and iii) (). Mantel tests were used to test the relations between pairwise taxonomic (IST) or phylogenetic distances (BST and ΠST) and geographic distances among the 9 transects using the R package vegan . A significant test with IST but not with BST or ΠST is expected under hypothesis (i).Finally, to assess the robustness of the results with respect to the taxonomic scale investigated, and possibly assess whether hypothesis (iv, compensatory effects between habitat filtering and competitive exclusion) might hold, partial randomization of the data between transects was performed on certain clades defined as species rich which were arbitrarily defined as families containing 10 or more sampled species. We also looked at Eudicot and Magnoliales clades. For each clade the coefficients described above were calculated under the 1s model (999 randomization of tree tips). This was done by using the spacodi.per.nodes function in the SpacodiR . […]

Pipeline specifications

Software tools Phylomatic, Mesquite, Phylocom, spacodiR
Application Phylogenetics