Neontology & Paleontology A unified framework for inferring phylogenies with fossils joint work with Tracy Heath & John Huelsenbeck Tanja Stadler Computational Evolution group — Department of Biosystems Science and Engineering [email protected] Combining molecular and fossil data ACACACCC Species1 ACACACCC Classic Species2 TCACACCT TCACACCT phylo- Species3 AAAGACTT AAAGACTT genetics Species4 ACAGACTT ACAGACTT 2 Combining molecular and fossil data ACACACCC Species1 ACACACCC Classic Species2 TCACACCT TCACACCT phylo- Species3 AAAGACTT AAAGACTT genetics Species4 ACAGACTT ACAGACTT Prior for each fossil calibration point Fossil Yang & Rannala (MBE, 2006) calibration Heath (Syst. Biol., 2010) Time 3 Combining molecular and fossil data ACACACCC Species1 ACACACCC Classic Species2 TCACACCT TCACACCT phylo- Species3 AAAGACTT AAAGACTT genetics Species4 ACAGACTT ACAGACTT Prior for each fossil calibration point Fossil Yang & Rannala (MBE, 2006) calibration Heath (Syst. Biol., 2010) Time Non-mechanistic models for speciation-extinction-fossilization Limitation Prior for each fossil to define Speciation and extinction rates cannot be estimated from trees with fossils 3 The birth-death model as a model for speciation, extinction and fossilization • stem age of a clade t λ μ • speciation rate λ Species • extinction rate μ Birth-death • fossil-observation rate ψ model • sampling probability ρ ψ Stadler (JTB, 2010) 4 The birth-death model as a model for speciation, extinction and fossilization • stem age of a clade t λ μ • speciation rate λ Species • extinction rate μ Birth-death • fossil-observation rate ψ model • sampling probability ρ ψ t Phylo- genetic trees complete phylogeny Stadler (JTB, 2010) 4 The birth-death model as a model for speciation, extinction and fossilization • stem age of a clade t λ μ • speciation rate λ Species • extinction rate μ Birth-death • fossil-observation rate ψ model • sampling probability ρ ψ t Phylo- genetic trees complete phylogeny reconstructed phylogeny Stadler (JTB, 2010) 4 The birth-death model as a model for speciation, extinction and fossilization • stem age of a clade t λ μ • speciation rate λ Species • extinction rate μ Birth-death • fossil-observation rate ψ model • sampling probability ρ ψ t Lik ( ) Phylo- genetic trees complete phylogeny reconstructed phylogeny Stadler (JTB, 2010) 4 Combining molecular and fossil data: Likelihood of the phylogeny 1 4λρ 4λρ p(y )q(y ) If ψ 2λ f f Lik ( ) = λ(1 − p0(x0)) q(x0) q(xi) q(zf ) i∈V f∈F Stadler (JTB, 2010) 5 Calibrating molecular phylogenies using the fossilized birth-death (FBD) model Heath, Huelsenbeck, Stadler (PNAS, 2014) 6 Calibrating molecular phylogenies using the fossilized birth-death (FBD) model Heath, Huelsenbeck, Stadler (PNAS, 2014) 6 Calibrating molecular phylogenies using the fossilized birth-death (FBD) model Heath, Huelsenbeck, Stadler (PNAS, 2014) 6 Calibrating molecular phylogenies using the fossilized birth-death (FBD) model Heath, Huelsenbeck, Stadler (PNAS, 2014) 6 Calibrating molecular phylogenies using the fossilized birth-death (FBD) model Heath, Huelsenbeck, Stadler (PNAS, 2014) 6 Calibrating molecular phylogenies using the fossilized birth-death (FBD) model Heath, Huelsenbeck, Stadler (PNAS, 2014) 6 Calibrating molecular phylogenies using the fossilized birth-death (FBD) model We sample all fossil attachments and λ, μ, ψ in an MCMC using our likelihood function (ρ=1) Heath, Huelsenbeck, Stadler (PNAS, 2014) 6 Simulation study Trees on 25 extant species were simulated with: Turnover: μ/λ = 0.5 Net diversification rate: λ-μ = 0.01 Fossilization rate: 0.1, Observation: 10% ψ = 0.01 Heath, Huelsenbeck, Stadler (PNAS, 2014) 7 Simulation results Heath, Huelsenbeck, Stadler (PNAS, 2014) 8 Simulation results Heath, Huelsenbeck, Stadler (PNAS, 2014) 8 More fossils yield smaller credible intervals Observation Heath, Huelsenbeck, Stadler (PNAS, 2014) 9 What about model violations in fossilization / observation? Heath, Huelsenbeck, Stadler (PNAS, 2014) 10 Biased fossil sampling Heath, Huelsenbeck, Stadler (PNAS, 2014) 11 Stratographic fossil sampling Heath, Huelsenbeck, Stadler (PNAS, 2014) 12 What about model violations in extant species sampling? Outgroup collapsed Deep nodes (young speciation events pruned) Heath, Huelsenbeck, Stadler (PNAS, 2014) 13 What about model violations in extant species sampling? Outgroup collapsed Deep nodes (young speciation events pruned) Heath, Huelsenbeck, Stadler (PNAS, 2014) 13 Data analysis: Dating the phylogeny of bears Heath, Huelsenbeck, Stadler (PNAS, 2014) 14 Summary All fossils can be used Calibrating No arbitrary priors for each calibration node phylo- genies Very robust towards model misspecification (biased fossil and extant species sampling) 15 Summary All fossils can be used Calibrating No arbitrary priors for each calibration node phylo- genies Very robust towards model misspecification (biased fossil and extant species sampling) Add Work in progress morpho- RevBayes logical Beast v2.0 information 15 Summary All fossils can be used Calibrating No arbitrary priors for each calibration node phylo- genies Very robust towards model misspecification (biased fossil and extant species sampling) Add Work in progress morpho- RevBayes logical Beast v2.0 information Model unifying processes leading to paleontological and Fossilized neontological data birth-death Allows to estimate speciation and extinction rates from model phylogenies with fossils 15 Phylogeny of Acknowledgements Paper: Heath, Huelsenbeck, Stadler (PNAS, 2014) DPPDiv: github.com/trayc7/FDPPDIV BEAST2: beast2.cs.auckland.ac.nz/ RevBayes: to come soon! A. Gavryushkina T. Heath A. Drummond J. Huelsenbeck.