Testing Phylogenetic Methods to Identify Horizontal Gene Transfer

Abstract

The subject of this chapter is to describe the methodology for assessing the power of phylogenetic HGT detection methods. Detection power is defined in the framework of hypothesis testing. Rates of false positives and false negatives can be estimated by testing HGT detection methods on HGT-free orthologous sets, and on the same sets with in silico simulated HGT events. The whole process can be divided into three steps: obtaining HGT-free orthologous sets, in silico simulation of HGT events in the same set, and submitting both sets for evaluation by any of the tested methods.Phylogenetic methods of HGT detection can be roughly divided into three types: likelihood-based tests of topologies (Kishino-Hasegawa (KH), Shimodaira-Hasegawa (SH), and Approximately Unbiased (AU) tests), tree distance methods (symmetrical difference of Robinson and Foulds (RF), and Subtree Pruning and Regrafting (SPR) distances), and genome spectral approaches (bipartition and quartet decomposition analysis). Restrictions that are inherent to phylogenetic methods of HGT detection in general and the power and precision of each method are discussed and comparative analyses of different approaches are provided, as well as some examples of assessing the power of phylogenetic HGT detection methods from a case study of orthologous sets from gamma-proteobacteria (Poptsova and Gogarten, BMC Evol Biol 7, 45, 2007) and cyanobacteria (Zhaxybayeva et al., Genome Res 16, 1099-108, 2006).

Prokaryotic systematics in the genomics era

“…The retained genes should then concatenated to maximize the phylogenetic signal and enhance the statistical support for branches in the tree inferred by the large dataset. Next, the individual genes should be the subject of another test, such as the Kishino-Hasegawa (KH), Shimodaira-Hasegawa (SH) or the approximately unbiased (AU) test (Poptsova 2009) on the supposition that the super-tree is the best. Crucially, when many genes are used in an analysis, it is necessary to account for the fact that different genes undergo different selective pressures hence the rate heterogeneity within sites may vary from gene to gene (Bevan et al 2007).…”

Inferring Horizontal Gene Transfer

“…Indeed, combining different parametric methods has been reported to significantly improve the quality of predictions [ 10 , 11 ]. Moreover, in the absence of a comprehensive set of true horizontally transferred genes, discrepancies between different methods [ 12 , 13 ] might be resolved through combining parametric and phylogenetic methods. However, combining inferences from multiple methods also entails a risk of an increased false positive rate [ 14 ].…”

A Phylogeny-Based Benchmarking Test for Orthology Inference Reveals the Limitations of Function-Based Validation

“…HGT, a prominent mechanism of genome evolution mainly in unicellular species, can be identified only through phylogenetic analysis [41], [42]. To the best of our knowledge, no existing orthology identification approach explicitly removes xenologs from the predicted orthologous groups.…”