How and Why Overcome the Impediments to Resolution: Lessons from rhinolophid and hipposiderid Bats

Foley, Nicole M.; Thong, Vu Dinh; Soisook, Pipat; Goodman, Steven M.; Armstrong, Kyle N.; Jacobs, David S.; Puechmaille, Sébastien J.; Teeling, Emma C.

To test whether exons perform better at resolving older nodes and if introns are better at resolving younger nodes, we compared the relationship between PP and height (referred to as node height and is defined as the average height of the clade across all trees in a BEAST run that support that clade) for intron and exon data sets. This analysis was carried out using the results of BEAST runs from analysis of the "exons -outliers removed" and "introns -outliers removed" data sets. It was not possible to compare directly the two outliers removed data sets (exons vs. introns) as the exon alignment was approximately three times larger (~7,000 bp vs.~2,500 bp), hence it would be expected that the exons should have more resolving power simply because there are more data. To overcome this, positions from the concatenated exon alignment were jack-knifed to form ten random exon data sets of 2,532 bp using custom R scripts. For each of the ten random exon data sets, jmodeltest indicated GTR+I+G was the best model. Each data set was analyzed in BEAST as described previously. We tested whether node height was related to PP using a general linear model with a binomial function and logit link. Differences in posterior probabilities for the same node between exons (average over the ten subsets) and introns were then tested in R using an asymptotic Wilcoxon Mann-Whitney rank sum test (Hollander and Wolfe 1999) . To investigate whether the results were specific to the jack-knifed data sets, we repeated the analysis using gene fragments. To obtain size comparable data sets, ten random combinations of four exon gene fragments