Influence of different modes of morphological character correlation on phylogenetic tree inference

Author(s): Thomas Guillerme, Abigail Pastore and Martin Brazeau

This repository contains all the code and data used in the manuscript Link to final published pdf will be here.

Data

This manuscript was based entirely on simulated data. Below are the procedures to repeat the data simulation. All the simulated data used in the manuscript is archive on Figshare (see below).

Simulating the data

The data simulation procedure is detailed in the paper and based on functions from the dispRity package. A tutorial to generate the "normal" matrix and modify it into the "minimised", "maximised" and "randomised" one is available here in Rmd or here in html.

Running the simulations

The simulations were ran on the Imperial College High Performance Computing clusters (2-3GHz clock rate). A tutorial for running the tree inferences specific to this cluster is available here in Rmd or here in html.

The script for generating the cluster jobs can be run using the set.runs.sh script. For example, use:

sh set.runs.sh 25t_100c_0001 MrBayes 8

For generating the MrBayes script for the 25t_100c_0001 chain (the first simulation for 25 taxa and 100 chains) for a 8 CPU node.

Comparing the trees

The trees where compared using the TreeCmp javascript from Bogdanowicz et al 2012. The script used for running the comparisons was CC_TreeCmp.sh. For example, use:

sh CC_TreeCmp -c 25t_100c_0001 -r 1

For comparing the minimised, maximised and randomised tree from the chain 25t_100c_0001 to the normal tree.

Archived data

The data used for the subsequent analysis is available on Figshare. The folder contains:

• Matrices containing all 1260 matrices for all number of taxa (25, 75, 150), all numbers of characters (100, 350, 1000) and all modifications (normal, minimised, maximised, randomised) replicated 35 times each. This folder also contains the "true" trees used to generate the matrices.
• Trees/ containing the consensus trees:
  • In Bayesian/ for the Bayesian inference trees.
  • In Parsimony/ for the maximum parsimony trees. Both contains the trees per number of taxa and characters (e.g. 25t_100c for 25 taxa and 100 characters)
• Comparisons/ containing the tree comparison scores: Containing again both Bayesian/ and Parsimony/ folders with their subfolders per taxa and characters for the comparisons to the normal tree (e.g. 25t_100c_cmp_norm) or to the randomised tree (e.g. 25t_100c_cmp_rand).

Analyses

The tables and figures present in the manuscript are all reproducible through the following scripts:

01 - Effect of character correlation

This script contains all the statistical analysis and results presented in the manuscript. This script is available here in Rmd or here in html.

02 - Full comparisons results

This script contains all the statistical analysis present in the manuscript (identical as in the one above) but allows to generate all the tables from the manuscript (including the descriptive tables in the supplementary materials and the testing of anova assumptions). This script is available here in Rmd or here in html.

03 - Figures

This script contains all the code to generate the figures that are in the manuscript or in the supplementary materials. This script is available here in Rmd or here in html.

Checkpoint for reproducibility

To rerun all the code with packages as they existed on CRAN at time of our analyses we recommend using the checkpoint package, and running this code prior to the analysis:

checkpoint("2018-04-19")

For reproducibility purposes, the output of devtools::session_info() used to perform the analyses in the publication is available here.