evalTrees
evalTrees trees.nwk > summary.txt
Evaluate properties of a collection of phylogenetic trees, typically representing samples from a posterior distribution. By default, summarizes univariate distributions of all pairwise distances between taxa in terms of median, mean, 95 pct credible interval, min, max, 25th and 75th percentiles and standard deviation. But can alternatively evaluate the distribution of log likelihoods using a specified alignment. If this alignment represents held-out data matching the data used to fit the model, the results can be interpreted as log pointwise predictive densities (LPPDs), a measure of model fit that can meaningfully be compared across different inference methods (e.g., Bayesian and ML). Can also evaluate the variability of topologies using –topology, or the branch-length accuracy of the trees relative to a reference tree using –branch-score.
Options
--help, -hPrint this help message and exit.
--model-fit, -f alignment.faEvaluate log likelihood of each sampled tree relative to the specified alignment. By default uses JC69 model but alternatives can be specified using –hky-kappa or –tree-model.
--hky-kappa, k <kappa>(for use with –model-fit) Use HKY85 model with specified value of kappa when evaluating log likelihood.
--tree-model, -m <file.mod>(for use with –model-fit) Use specified PHAST-formatted tree model when evaluating log likelihood.
--crispr, -c(for use with –model-fit) Use CRISPR model rather than DNA model. In this case, argument to –model-fit will be expected to be a *.tsv-formatted mutation matrix.
--topology, -t <reference.nwk>Instead of pairwise distances between taxa, consider the distribution of topological distances of all sampled trees from a given reference tree topology.
--branch-score, -b <reference.nwk>Evaluate the branch-score (Kuhner-Felsenstein) distance of the sampled trees to a given reference tree. Unlike –topology, the lengths of all edges (including terminal edges) contribute, so this measures branch-length as well as topological agreement. Reports the distribution of per-sample distances (as for –topology) followed by two additional lines: “Point (posterior-mean-tree) BSD”, the distance between the reference and the posterior-mean split-length vector – a point estimate of branch-length accuracy that is insensitive to posterior dispersion; and “Reference tree length”, the total branch length of the reference (useful for normalization).
--entropy, -eCompute uncertainty measures for the tree distribution. Reports the number of unique tree topologies and three uncertainty measures: (1) “Split entropy”: sum of Bernoulli entropies over all non-trivial splits weighted by their posterior inclusion probabilities; (2) “Topology entropy”: Shannon entropy over distinct topologies, -sum p(tau) log p(tau); (3) “Mean branch-length variance per branch”: topology-frequency-weighted mean of sample variances of log branch lengths, averaged over branches.
--rf-matrix, -r <file.tsv>Print an upper-triangular, tab-separated matrix of Robinson-Foulds distances between all input trees to the specified file. Rows and columns are labeled by input order.
--rf-mds, -M <file.tsv>Compute classical multidimensional scaling of the Robinson-Foulds distances between input trees and print the unscaled two-dimensional coordinates to the specified tab-separated file. Can be used together with –rf-matrix to produce both outputs in one run.