Computational protocol: The Braincase of Eocaecilia micropodia (Lissamphibia, Gymnophiona) and the Origin of Caecilians

Similar protocols

Protocol publication

[…] Because the braincase was the focus of the current study, ancestral character state reconstructions of the 34 caecilian braincase characters from Maddin et al. were performed in order to assess whether or not the condition present in Eocaecilia micropodia is representative of the plesiomorphic condition of the braincase for Apoda. The state of each character at the base of Apoda was estimated using maximum parsimony and maximum likelihood (Mesquite v.2.72; ), and Bayesian inference (BayesTraits v.1.0; ). For the maximum parsimony and maximum likelihood approaches, the maximum clade credibility tree from the analysis of Maddin et al. was used as the topology upon which character evolution was reconstructed. For the Bayesian estimation a sample of 1000 post-burn-in trees generated in the Bayesian analysis of phylogeny from Maddin et al. were imported into BayesTraits, along with an input file consisting of the character states associated with each species for the thirty-four braincase and stapes characters. The Multistate and MCMC options were selected for the analyses. Each ancestral character state reconstruction analysis was run for 1 million iterations, at which point the harmonic means of the log likelihoods were observed to reach stationarity. The run was sampled every 1,000 iterations, after a burn-in period of 100,000 iterations. The rate of deviation parameter (ratedev) was adjusted to obtain a recommended level of acceptance (20–40%, in this case, using a ratedev of 5), and the reverse jump hyperprior was set to exponential on an interval from 0 to 30, as per the program’s recommendations. The Fossil Node function was applied to each character to test hypotheses of which state is more likely to occur at the node of interest, in this case the base of Apoda. This function ‘fossilizes’ the state at this node and indicates the probability of that state being present at that node, based on the distribution of states in the terminal taxa. State 0 was first fossilized, followed by state 1 (then state 2, etc., if applicable). To determine which state is more likely, a Bayes Factor test was applied. This was accomplished by subtracting the harmonic mean of the log likelihoods after 1 million iterations for state 1 from the harmonic mean of the likelihoods for state 0, and multiplying this value by 2 . In general, positive values are taken as favouring the first model, but are not necessarily statistically robust. Values greater than 2 are considered to support the first model, values greater than 5 are considered to strongly support the first model, and values greater than 10 to very strongly support the first model . The results of the three approaches were compared for congruence. […]

Pipeline specifications

Software tools Mesquite, BayesTraits
Application Phylogenetics