Computational protocol: A multi-locus inference of the evolutionary diversification of extant flamingos (Phoenicopteridae)

Similar protocols

Protocol publication

[…] Sequences were aligned within Sequencher 4.8 and alignments were confirmed within MEGA 5.2 [] using ClustalW with default settings. Pairwise genetic distances were calculated in MEGA 5.2 using the default Maximum Composite Likelihood substitution model except for COI, which was calculated using Kimura 2-Parameter. Maximum likelihood phylogenetic analyses were carried out within MEGA 5.2 and the loci were concatenated. Unless the dataset comprised only coding genes (ZENK, COI and cyt b), each dataset was treated as noncoding; optimal substitution models were found within the hierarchical BI and ML framework in MEGA 5.2 (Table ) and the analysis used a nearest-neighbor-interchange heuristic search. Bootstrap support was estimated from 500 replicates using the same settings. Primary phylogenetic analyses using Bayesian inference were carried out in MrBayes 3.2.1 []. Datasets were fully partitioned based on the optimal substitution models estimated within MEGA 5.2. Each search comprised two concurrent runs of four chains each for 10,000,000 generations sampled every 1,000 generations with the first 1,000,000 generations discarded as burn-in. The results of each analysis were tested for convergence of phylogenetic signal in Tracer 1.5 [], where estimated sample size (ESS) values >200 were treated as reliable signal.The primary analysis employed all 10 nDNA and two mtDNA loci for one individual per species. All secondary analyses employed the same parameters as the primary analysis except where noted. To test for the presence of intraspecific variation, a series of single-locus analyses were carried out using two to three sequences per flamingo species for five loci (Additional file ). To recover individual gene trees, each locus was analyzed independently. To test the relative weight of each locus to the combined dataset, a series of analyses were run excluding a single gene at a time. To test for conflicting signal between the nDNA and mtDNA datasets, each was analyzed independently. To test for the effects of over-partitioning of the Bayesian inference analysis, a Bayesian analysis was carried out treating the total dataset as non-partitioned (as in the ML analysis). A Bayes factor analysis was carried out in Tracer 1.5 to compare the strength in phylogenetic signal of the fully partitioned and non-partitioned BI treatments.*BEAST 1.8.0 [] was used to test for discordance between individual gene trees and the species tree. The dataset was fully phased by individual and gene. Loci were allowed unlinked substitution models, clock models and partition trees. The optimal available substitution model was used (Table ) and the 3 coding loci were partitioned by codon position. Each locus was allowed a lognormal relaxed clock with uncorrelated rates. Mean clock rate priors (ucld.mean) were allowed uninformative uniform distributions (Initial value = 0.05; Upper = 1.0E100; Lower = 0.001). A Yule speciation process was assumed. No fossil node calibrations were used. The results were tested for convergence of phylogenetic signal in Tracer 1.5. [...] Divergence time estimation analyses were carried out using BEAST 1.8.0 []. Loci were allowed unlinked substitution models (Table ) and the 3 coding loci were partitioned by codon position. A lognormal relaxed clock with uncorrelated rates was assumed and a Yule speciation process was employed. The clock mean (ucld.mean) was allowed an uninformative uniform prior distribution (Initial value = 0.05; Upper = 1.0E100; Lower = 0.001). Taxa were divided into sets representing grebes and Mirandornithes; monophyly of these sets was not enforced. Two nodes were fossil calibrated following the guidelines suggested by Parham et al. []. The divergence of extant grebes was calibrated at 8.7 mya based on Thiornis sociata, a grebe from the late Miocene of Spain [,]. The divergence of flamingos and grebes was calibrated at 32.6 mya based on Adelalopus hoogbutseliensis, a stem phoenicopteriform from the earliest Oligocene of Belgium []. Calibrations were allowed lognormal distributions with offset = calibration age and standard deviation = 1.0. See Table for the data included in each divergence analysis. All analyses were run twice for 100 million generations sampled every 1000 with burn-in of 1 million. Convergence of phylogenetic signal was tested for in Tracer 1.5. […]

Pipeline specifications

Software tools Sequencher, MEGA, Clustal W, MrBayes, BEAST
Application Phylogenetics
Organisms Phoenicopterus roseus