Phenotypic evolution: the emergence of a new synthesis

Phenotypic evolution: the emergence of a new synthesis

Phenotypic evolution: the emergence of a new synthesis Stevan J. Arnold Oregon State University Outline Synthesis in evolutionary biology then and now Simpson (1944) & the ongoing synthesis in evolutionary quantitative genetics Two examples of the ongoing synthesis (Estes & Arnold 2007, Uyeda et al. 2011)

Conclusions & perspectives from the two studies Some general lessons about synthesis in evolutionary biology Synthesis in evolutionary biology Cumulative number of citations of 57 influential books as a function of time 200000 Citations

150000 100000 50000 0 1850 1870

1890 1910 1930 Year 1950 1970

1990 2010 Synthesis 1930-32 (a) R A Fisher 1930 The genetical theory of natural selection..12,618 citations (b) S Wright 1931 Evolution in Mendelian populations... 5,493 (c) J B S Haldane 1932 The causes of evolution. 1,463 Synthesis 1937-50 T Dobzhansky 1937 Genetics and the origin of species.. 4,591 citations

R Goldschmidt 1940 The material basis of evolution.. 1,009 E Mayr 1942 Systematics and the origin of species . 4,380 J Huxley 1942 Evolution, the modern synthesis. 1,891 G G Simpson 1944 Tempo and mode in evolution 1,684 I I Schmalhausen 1949 Factors of evolution .... 841 G L Stebbins 1950 Variation and evolution in plants 3,506 Dobzhansky Goldschmidt Mayr

Huxley Simpson Schmalhausen Stebbins Synthesis in evolutionary biology An ongoing process since 1859, especially now! Cumulative number of citations of 57 influential books as a function of time 200000 Citations

150000 100000 50000 0 1850 1870

1890 1910 1930 Year 1950 1970

1990 2010 Simpsons 1944 synthesis Population genetics meets paleontology evolution in deep evolutionary time Reliance on case studies Qualitative use of theory Use of graphical models (e.g., adaptive

landscape for phenotypic traits) Simpsons concept of quantum evolution Simpsons concept of quantum evolution Ongoing synthesis in evolutionary quantitative genetics Quantitative genetics provides a theoretical framework with direct connections to data Key concepts rendered in statistical terms

Mega-data sets reveal evolutionary patterns Test alternative models in ML framework Two examples of ongoing synthesis Suzanne Estes & S J Arnold 2007 Resolving the paradox of stasis: models with stabilizing selection explain evolutionary divergence on all timescales.

American Naturalist Suzanne Estes Two examples of ongoing synthesis Josef C Uyeda, Thomas F Hansen, S J Arnold & Jason Pienaar 2011 The million-year wait for macroevolutionary bursts. PNAS USA

Josef Uyeda Thomas Hansen Jason Pienaar The approach Make data and theory communicate (Plot your data!) Compile abundant, high quality data (necessarily univariate) Compile a priori estimates of key parameters:

population size, inheritance, selection Use the most powerful stochastic models of phenotypic evolution, cast in terms of key parameters Confront the models with data (cross-check with parameter estimates) A priori estimates of key parameters Heritability n=580 D Roff, pers com

Derek Roff Stabilizing selection Distance to optimum n=355 n=197

Kingsolver et al. 2001 Joel Kingsolver Kingsolver et al. 2001 A short digression to talk about stochastic models of phenotypic evolution If replicate lineages obey the same stochastic rules,

we can statistically characterize the distribution of trait means of those replicates at any generation in the future. Joe Felsenstein Russ Lande Mike Lynch A short digression to talk about

stochastic models of phenotypic evolution For example, in the case of drift with no selection, the mean at a particular generation is the sum of two parts: (a) the mean in the preceding generation (b) deviation due to parental sampling, a normally distributed variable with zero mean and a variance equal to G/Ne , where G is genetic variance and Ne is effective population size

A short digression to talk about stochastic models of phenotypic evolution If drifting replicate lineages obey the same stochastic rules, we can statistically characterize the distribution of lineage trait means at any generation, t, in the future. In this particular case, the replicate trait means will be normally distributed with zero mean and a variance equal to tG/Ne.

Lineage mean A simulation of a single lineage evolving by drift Time (generations) A simulation of 100 lineages evolving by drift Lineage mean

99% confidence limits Time (generations) Testing models with the Gingerich data The data (sources, pattern) The models (drift, models with a stationary optimum, models with a moving optimum) Conclusions

Estes & Arnold 2007 Testing models with the Gingerich data The data (sources, pattern) The models (drift, models with a stationary optimum, models with a moving optimum) Conclusion Andrew Hendry Philip Gingerich

Michael Kinnison Estes & Arnold 2007 Testing models with the Gingerich data The data (sources) Longitudinal data: 2639 values for change in trait mean over intervals ranging from one to ten million generations; 44 sources, time series. Traits: size and counts; dimensions and shapes of shells, teeth, etc.; standardized to a common scale of within-population,

phenotypic standard deviation. Taxa: foraminiferans ceratopsid dinosaurs. Estes & Arnold 2007 A short digression to talk about the data plots Mean body size at generation 0 = 100 mm Mean body size at generation 100 = 150 mm Average within-population std dev in body size = 10 mm Divergence = 150 mm - 100 mm = 50 mm or 5 sd

Interval = 200 = 102 generations A short digression to talk about the data plots Mean body size at generation 0 = 100 mm Mean body size at generation 100 = 150 mm Average within-population std dev in body size = 10 mm Divergence = 150 mm - 100 mm = 50 mm or 5 sd Interval = 200 = 102 generations Testing models with the Gingerich data The data (pattern): 99% confidence ellipse

Estes & Arnold 2007 Testing models with the Gingerich data (Estes & Arnold 2007) The data (pattern): 6 within-pop pheno sd Estes & Arnold 2007 When models confront the data, they can fail in three ways

1. Under-prediction: no points here Estes & Arnold 2007 When models confront the data, they can fail in three ways 2. Blowout: lots of points here Estes & Arnold 2007 When models confront the data, they can

fail in three ways 3. Fails parameter cross-check: requires unrealistic values Estes & Arnold 2007 Testing models with the Gingerich data Conclusion: When representatives of the entire family of existing stochastic process models confront the data, only a single model is left standing. Drift (Brownian motion) Stationary optimum (OU)

Fluctuating optimum (Brownian motion or white noise) Moving optimum (with white noise) Peak shift (drift from one optimum to another) Genetic constraints with any of the above Displaced optimum model Estes & Arnold 2007 Displaced optimum model Response of one lineage mean

Model of peak movement Lande 1976 Displaced optimum model Lineage mean Multiple lineages chasing displaced optima could easily fill an adaptive zone

Time (generations) Lande 1976 Testing models with the Uyeda et al data The data (sources, pattern) The models: white noise fluctuation of the trait mean combined with three models of moving optima (Brownian motion, singleburst, multiple-burst) Conclusions

Uyeda et al 2011 Testing models with the Uyeda et al. data The data (sources) Size-related traits: over 8,000 data points from 206 studies. Three sources: (i) microevolutionary time series, (ii) fossil time series, (iii) data from time-calibrated trees.

Vertebrate taxa: mammals, birds, squamates. Uyeda et al 2011 A hypothetical data point on the new plotting axes 65% change in body size The Blunderbuss Pattern

Testing models with the Uyeda et al data The data (sources, two parts to the barrel of the blunderbuss) The models (white noise = the base of the barrel, models with moving optima = the flared end of the barrel: Brownian motion, single-burst model, multiple-burst model Conclusions

Uyeda et al 2011 Modeling strategy Account for the long barrel of the blunderbuss with a surrogate process (white noise fluctuation of the lineage mean about the trait optimum) Compare 3 alternative models to account for the flared end of the blunderbuss (Brownian motion and two descendants of the displaced optimum model). The 2 descendants: single- and multiple burstmodels.

Uyeda et al 2011 Simulations of the single-burst model Lineage mean (peak movement, evolution of the lineage mean) A single lineage Lineage mean

Time (generations) Multiple lineages Time (generations) Uyeda et al. 2011 Simulation of the multiple-burst model (peak movement, evolution of the lineage mean)

Lineage mean A single lineage Time (generations) Uyeda et al. 2011 Model comparisons Model

White noise (WN) only Brownian motion + WN Single-burst + WN Multiple-burst + WN White noise parameter estimate ( ) 0.20 0.11 0.10

0.10 AIC* 2940.53 7877.97 9018.0 9142.54 Model comparisons Model

White noise (WN) only Brownian motion + WN Single-burst + WN Multiple-burst + WN White noise parameter estimate ( ) 0.20 0.11 0.10

0.10 AIC* 2940.53 7877.97 9018.0 9142.54 Multiple-burst model: parameter estimates White noise

distribution (dashed) Burst size distribution (solid) Burst timing distribution (mean time between bursts = 25 my) Probability

Probability Conclusions & perspectives from the two studies Micro- and meso-evolution is bounded. What is the best model of that bounded evolution? Evolutionary bursts are rare but increasingly inevitable in deep evolutionary time. Is the blunderbuss pattern general? Are invasions of new adaptive zones responsible for

evolutionary bursts and hence the flared barrel of the blunderbuss? What triggers those bursts/invasions? Synthesis in evolutionary biology An ongoing activity since 1859 Contention and bickering is normal To synthesize, we need to bridge between fields Data should talk to theory & vice versa An extraordinary burst of synthesis is

happening right now! What about your synthesis? Acknowledgements Research collaborators: Suzanne Estes, Josef Uyeda, Thomas Hansen, Jason Pienaar, Phil Gingerich, Andrew Hendry, Michael Kinnison, Russell Lande, Adam Jones, Reinhard Brger and all of you! NESCent course collaborators: Joe Felsenstein, Trudy Mackay, Adam Jones, Jonathan Losos, Luke Harmon, Liam Revell, Marguerite Butler, Josef Uyeda, Matt Pennell

NSF OPUS program: Mark Courtney Editor/publisher: Trish Morse, Andy Sinauer

Recently Viewed Presentations

  • Metadata Topics Cardiff, Feb 2009

    Metadata Topics Cardiff, Feb 2009

    The Committee for the Implementation of CFI (CFI2) is overseeing the development of tools to manage the information surrounding the use of data and the RDC Intranet will be a valuable mechanism for sharing this information. Three projects are being...
  • William Paterson University Purchasing Department Requisition ...

    William Paterson University Purchasing Department Requisition ...

    Dept. Bid Waiver. Public Advertised Bid or Bidding Exception* or Board of Trustees Bid Waiver (Cumulative by fiscal year) Applicable for all orders and change orders. * Bidding requirement is satisfied by using: State contract, GSA contract Schedule 70 or...
  • Les contrats des service reliés à la sécurité informatique ...

    Les contrats des service reliés à la sécurité informatique ...

    Les contrats de services reliés à la sécurité informatique: éviter les pièges Conférence Insight : Sécurité informatique Les 14 et 15 janvier 2008
  • Do Now: Scan the information/graphics for each of

    Do Now: Scan the information/graphics for each of

    Do Now: Scan the information/graphics for each of the three food guides. Write one "understanding" in English about each of the guides. Ex: I understand that Americans value fruits and vegetables because they make up 50% of the plate.
  • Project: Chemical Compounds

    Project: Chemical Compounds

    Chemicals Compound Project OBJECTIVE: Students will become better acquainted with the writing and naming of chemical compounds and the use of chemicals in everyday life in foods and products that are used in the home.
  • Welcome to Maastricht University School of Business and

    Welcome to Maastricht University School of Business and

    webmail.maastrichtuniversity.nl. Professors/Tutors use only this UM mail address to communicate with you personally. Address book is accessible via the 'to' button. Make sure you receive all messages: install your student e-mail account on your smartphone and/or tablet. UM Student mail
  • In the Clinic Depression  Copyright Annals of Internal

    In the Clinic Depression Copyright Annals of Internal

    A second review of 19 controlled trials on adults older than 65 years supported this recommendation for methyl-phenidate; however, dosing, when to initiate therapy, and how to monitor side effects remains unclear (44).
  • High Risk Pregnancy - WordPress.com

    High Risk Pregnancy - WordPress.com

    High-Risk Pregnancy. Jeopardy to mother, fetus, or both. Condition due to pregnancy or result of condition present before pregnancy. Higher morbidity and mortality. Risk assessment with first antepartalvisit and each subsequent visit . Risk factors