Computational protocol: Tempo and Mode of the Evolution of Venom and Poison in Tetrapods

Similar protocols

Protocol publication

[…] We estimated ancestral states for each of our categories using Bayesian stochastic mapping [] on each major clade separately. We ran these analyses in the R package phytools [] based on an “ARD” model, in which gains and losses can occur at different rates. The prior on the root state was estimated using transition matrices within the function. We generated 1000 stochastic maps for amphibians, mammals, and squamates which were used to reconstruct ancestral states. For birds, we instead simulated 10 stochastic maps on each of the 50 phylogenies. This allowed us to incorporate the uncertainty in this set of trees, but the size and number of the bird phylogenies exhausted available computer memory and so we based our inference on the smaller number of 500 maps in contrast to the 1000 used for other groups. Nevertheless, this should still give adequate information for inference of basic macroevolutionary patterns.Since the size of the four tetrapod clades examined herein obscures detail when ancestral states at all nodes are plotted, we instead visualised the results by plotting the probable shifts between different character states on each tree. For this we plotted pie charts on branches in which the probability of a gain or loss of a trait was highest, giving a clearer picture of how and where each category has changed over evolutionary time. [...] To estimate transition rates while controlling for potential effects of each category of toxic weaponry on diversification [], we used binary state speciation and extinction (BiSSE) models []. We fit these models in the R package diversitree which incorporates extensions to the original BiSSE model to account for incomplete sampling []. We note that BiSSE models have been criticised as tools for investigating diversification rates, particularly extinction rates [], but we did not interpret diversification parameters here. Instead, we merely included them in the model to account for relationships between the traits and diversification while only extracting estimates of transition rates from the models. We first fit BiSSE models to each category on each phylogeny using maximum likelihood (ML), then used the ML estimates as priors in a Markov chain Monte Carlo (MCMC) analysis in order to incorporate uncertainty in parameter estimates. We initially ran a Markov chain for 1000 iterations to optimise step size and then used this optimised value in the final MCMC run of 15,000 steps. We conservatively discarded the first 5000 samples as burn-in and, therefore, used the last 10,000 for inference. For our bird phylogenies, we ran an MCMC for 5200 steps over each of the 50 trees, discarding the first 5000 as burn-in from each run, and combined the 200 posterior samples from each tree for inference (again giving 10,000 samples in total). Frequency plots of the posterior distributions from these analyses are provided in . […]

Pipeline specifications

Software tools Phytools, Diversitree
Application Phylogenetics