Thus, rather than assuming an identical pre-determined fragment size for all fragments and samples, we estimated the fragment size distribution for each sample from the empirical sequencing data.
These tests using BLAST were employed as a complement to the primary results on barcode loci derived from the empirical comparative sequence data set.
However, other than simulation studies, there are few means of testing the validity of this assumption using empirical gene sequence data.
In particular, we increased the among-gene variance of the simulated data by incorporating variation in models of molecular evolution, model parameter values, and gene tree topologies derived from our empirical nuclear sequence data set (see below).
Phylogenetic analyses of the mitochondrial and empirical concatenated nuclear sequence data sets were performed under ML, MP and Bayesian Inference.
Results: With empirical protein-sequence data, Bayesian posterior probabilities provide more-generous estimates of subtree reliability than does the nonparametric bootstrap combined with maximum likelihood inference, reaching 100% posterior probability at bootstrap proportions around 80%.
For empirical protein-sequence data that might reasonably be encountered in biological research, then, both gamma-corrected ML and gamma-corrected Bayesian inference perform well in recovering the correct topology.
With empirical protein-sequence data, Bayesian posterior probabilities provide more-generous estimates of subtree reliability than does the nonparametric bootstrap combined with maximum likelihood inference, reaching 100% posterior probability at bootstrap proportions around 80%.
However, whereas for empirical protein-sequence data (and nucleotide-sequence data: see references cited immediately above) PPs tend to "saturate", i.e. reach 100% at BP values less than 100% (here around 80%), with our simulated data the relationship between BP and PP resembles a smooth curve reaching 100% PP only at BP greater than 99%.
