嘉宾（Guest）：Dr. Huw A. Ogilvie 

讨论主题（Topic）：Accounting for gene tree discordance in total evidence phylogenetics 

时间（Time）：2019.1.10（周四）下午 2:30 — 3:30 

地点（Location）： 南楼321室 

内容简介：Recombination shuffles the evolutionary history of genomes with every generation. As a consequence, while a single phylogenetic tree is sufficient to represent the evolutionary history of multiple genomes at a single locus, different loci will follow different gene trees. Only some, or even none, of the gene trees will be identical to the species tree that represents the evolutionary history of the corresponding species. Theoretical and simulation studies have shown that accurate inference of species trees must account for gene tree discordance. 

　　The multispecies coalescent (MSC) is the most popular model which accounts for such discordance, and has been implemented in packages like BPP, *BEAST and StarBEAST2. These methods are widely used with neontological data (i.e. genomic sequences from extant species), but were until now not applicable to paleontological data (i.e. the morphology of extinct species). Past analyses of combined neontological and paleontological data, also known as total evidence, have therefore employed methods which may not have inferred accurate species trees.

　　To address this situation, we have extended the MSC to incorporate morphological and paleontological data for total evidence analyses. Our model incorporates the fossilization-birth-death (FBD) process to model the pattern of speciation and extinction where extinct species have been sampled, and also incorporates the Mk substitution model of morphological evolution. Because our new model integrates the MSC model and FBD process, we call it "MSC-FBD".

　　Using simulation we have shown that the MSC-FBD model is able to accurately estimate species trees from total evidence data sets. However by applying it total evidence data from Caninae, the subfamily of Canindae which includes all extant species of dogs and foxes, we have shown that the Mk model applied to real data can produce erroneous results because it gives undue weight to highly correlated characters. We hope to address this shortcoming in future work.

报告人信息： 

　　Dr. Huw A. Ogilvie finished his PhD in 2018 supervised by Prof. Craig Moritz (Australian National University) and Prof. Alexei Drummond (University of Auckland). Currently, he is a postdoc and lecturer in Prof. Luay Nakhleh’s lab in Rice University. He has been productively working on phylogenetic methods of inferring species evolutionary histories while taking advantage of multilocus genomic sequences. For more information, please visit https://www.cs.rice.edu/~ogilvie/