Computational protocol: Plate tectonics drive tropical reef biodiversity dynamics

Similar protocols

Protocol publication

[…] We reconstructed the phylogenetic relationships of species occurring in tropical reefs of the Labridae family together with outgroups according to previous phylogenies. Molecular sequences from published sources were downloaded on GenBank focusing on the most frequent loci (). Those loci were then concatenated to perform a supermatrix approach for phylogenetic reconstruction with sufficient taxon overlap for each loci. We used four mitochondrial gene regions (non-coding: 12S and 16S, coding: Cytb and CO1) and three nuclear genes (RAG2, S7 II and tmo4c4), getting data for 55% of known species. Each marker data set was independently aligned using the algorithm MUSCLE with standard settings and alignments were then manually edited using MEGA 6 (ref. ). We tested substitution models using a maximum likelihood approach with a random starting tree for each gene alignment and we selected the best models using the BIC criterion on Jmodeltest 2.1.6 (refs , ; ). Alignments were afterwards concatenated in a supermatrix using SequenceMatrix. We then analysed the supermatrix using a Maximum Likelihood approach via the algorithm RAxML (ref. ) implemented in RAxMLGUI (ref. ). We performed 10 parallel runs to fully explore tree space and to avoid being retained on a local optimum. We then selected the best tree using likelihood scores. To estimate node ages, we used the penalized likelihood method on the best likelihood tree. We then used the ultrametric tree as a starting tree for Bayesian inference of tree topologies and node ages using Markov chain Monte Carlo (MCMC) in BEAST v1.8.1 (ref. ). We ran six independent MCMC, 40 × 106 steps long under individual gene models previously selected for each reconstruction. We used the same fossil and biogeography based calibrations of Cowman and Bellwood as we had no update in fossil calibrations since then, and there is no fossil data available to our knowledge for the additional families (). We checked for the convergence and stationarity of each independent run using Tracer v1.6 (ref. ). We combined the six independent runs for each data set (after removing an appropriate burn-in) using LogCombiner v1.8.1 (ref. ) to reach an effective sample size above 200 for all our estimates and we extracted the maximum clade credibility tree for combined tree sets using TreeAnnotator v1.8.1 (ref. ). The resulting time-calibrated phylogeny () is consistent with previously published phylogenies for the Labridae, with a few differences. Overall, the node age estimates appear to be 5–10 million years older than previously estimated (). In addition, few nodes showed low posterior probabilities (). [...] To compare to the simulated diversification rate to empirical patterns, we estimated extinction and speciation rates shaping diversification for the coral family Acroporidae using fossils occurrences following the method (PyRate) proposed by Silvestro et al. This method includes the probability of preservation and sampling of fossils to estimate the lifespan of each lineage. We first removed all undetermined taxa from the data set. Due to the lack of fossil occurrences before 60 Ma, we only estimated diversification rates during the last 60 Myr. We first ran 10,000,000 BDMCMC generations both assuming constant and Gamma distributed preservation rate to test the heterogeneity of preservation rate. We then generated 10 randomization of the age of fossils occurrences and performed 10 independent BDMCMC analyses to estimate the diversification rates through time. We examined the posterior samples using Tracer v1.6 (ref. ) and combined it after removing an appropriate burn-in to get a good estimation of speciation and extinction rates.Because missing extinct lineages in the most ancient section of the phylogeny of Labridae preclude any diversification estimate in the Cretaceous from phylogenies, we computed the rate of diversification from a phylogeny containing 18 tropical reef fish families (that is, Labridae including Scaridae, Pomacentridae, Chaetodontidae, Acanthuridae, Haemulidae, Balistidae, Carangidae, Serranidae, Lutjanidae, Sparidae, Caesionidae, Holocentridae, Mullidae, Muraenidae, Tetraodontidae, Lethrinidae and Siganidae). We obtained this phylogeny by pruning a time-calibrated phylogeny for 7,822 extant fish species. These families were selected as the most representative reef fish families, that is, they are abundant and speciose on tropical reefs. These families are well sampled in the phylogeny of Rabosky et al. with more than 80% of known genus. As missing species may influence the estimation of diversification rates, we grafted those species on the pruned phylogenetic tree (1,291 missing species out of 2,224) based on published phylogenies for these families, supplemented by taxonomic information from fish identification guides and FishBase (www.fishbase.org). Specifically, new tips representing unsampled species were added to direct sister species when information allowed or to the base of the clade representing its genus. Using this empirical phylogeny, we estimated diversification rates based on an evolutionary model under a birth-death-shift process. The birth-death-shift model uses a likelihood approach to estimate speciation and extinction rates like in a simple birth-death process but it also allows changes of those parameters through time if we force diversification shifts to happen at given times. For a given number of shifts, we simultaneously estimated time at which shifts occur, and speciation and extinction rates within each time interval between diversification shifts. We calculated those parameters for several numbers of shifts and we compared the resulting models using the AIC. The method proposed by Stadler is robust to phylogenetic uncertainty due to polytomies. Finally, we extracted the simulated values of diversification rate expected under the parapatric and sympatric models by computing for each time step the speciation rate minus the extinction rate and we compared the observed diversification rates with those obtained from the observed phylogeny for fishes and fossils for corals (). […]

Pipeline specifications

Software tools MEGA, jModelTest, Sequence Matrix, RAxML, raxmlGUI, BEAST, PyRate
Databases FishBase
Application Phylogenetics