Modeling Character Evolution with Phylogenetic Uncertainty

April M. WrightMatthew C. Brandley

Kathleen LyonsDavid M. Hillis

Is it possible to re-evolve the egg?

● Viviparity has evolved many times from oviparity

John Gould via WikiMedia Commons David Hillis

Is it possible to re-evolve the egg?

● Viviparity has evolved many times from oviparity

Does the reverse ever occur?

Oviparity as a Dollo Trait

Complex characters, once lost, are unlikely to re-evolve

Oviparity as a Dollo Trait

● Lost once in mammals● Never re-evolved

Warren et al, Nature

Oviparity as a Dollo Trait

● Lost once in mammals○ Never re-evolved

● Transition to viviparity observed almost 30 times in fish○ Evidence of re-evolution of oviparity?

Oviparity

Various forms of viviparity&

Dulvy and Reynolds

Oviparity as a Dollo Trait

● Lost once in mammals○ Never re-evolved

● Transition to viviparity observed almost 30 times in fish○ Evidence of re-evolution of oviparity?

● Transition to viviparity observed over 100 times in squamates○ Has oviparity re-evolved?

Oviparity as a Dollo Trait

● Has been proposed as a Dollo trait○ Previous phylogenetic analyses have placed

oviparous taxa within clades of viviparous taxa

Oviparity as a Dollo Trait● Has been proposed as a Dollo trait

○ Previous phylogenetic analyses have placed oviparous taxa within clades of viviparous taxa

○ Suggests possibility of reversals

Kris Kendall San Diego Zoo

Oviparity as a Dollo Trait

● Previous phylogenetic analyses have placed oviparous taxa within clades of viviparous taxa

● Comparative work by Fenwick et al. and Lynch and Wagner have supported reversibility in this trait

Should reversal surprise us?

● Several types of viviparity represented in reptiles

Should reversal surprise us?● Several types of viviparity represented in

reptiles

Mark Stevens via Wikimedia CommonsToby Hudson via Wikimedia Commons

Should reversal surprise us?● Several types of viviparity represented in

reptiles

Dutta and Medhi Dr. Anne Fawcett

A great test dataset

● Pyron and Burbrink assembled a 8000-taxon dataset on parity mode for extant squamates

A great test dataset

● Pyron and Burbrink assembled a 8000-taxon dataset on parity mode for extant squamates

● Coupled with a 4200-taxon tree○ 3950 taxa overlap

A great test dataset

● Their analysis:○ Fit a probabilistic model of character change to

these data○ Estimated ancestral states○ Concluded that the root state of the squamate tree

was viviparous, with strong statistical support

An ‘open’ question

● A maximum likelihood tree is a point estimate

● The root state of squamates had previously been thought to be oviparous, based on tetrapod phylogeny

Oviparity

Oviparity and viviparity

Viviparity

An ‘open’ question

● A maximum likelihood tree is a point estimate

● The root state of squamates had previously been thought to be oviparous, based on tetrapod phylogeny

● And a lovely, open dataset

A tour of the data set

A tour of the data set

Snakes

Iguanids

Anguimorphs

LacertidsSkinks

Geckos

● ~Half of oviparous species represented

● ~60% of viviparous

● 85% percent of overall extant Squamates

How much does uncertainty in the tree affect ancestral state

reconstruction?

Our approach

● Estimate ancestral states across a bootstrap sample of trees○ Estimate trees in Examl○ Time-scale trees using treePL

Our approach

● Estimate ancestral states across a bootstrap sample of trees○ Estimate trees in Examl○ Time-scale trees using treePL

● Fit a probabilistic model of character evolution to each tree in sample

Our approach

● Estimate ancestral states across a bootstrap sample of trees○ Estimate trees in Examl○ Time-scale trees using treePL

● Fit a probabilistic model of character evolution to each tree in sample

● Visualize the uncertainty in model parameters and ancestral states

Bootstrap sample

● Trees generally very similar in the major groups

Snakes

Iguanids

Anguimorphs

LacertidsSkinks

Geckos

Bootstrap sample

● Trees generally very similar in the major groups

● Most variation among tips

Fitting a model● Full BiSSe model

○ Two speciation parameters

○ Two extinction parameters

○ Two transition rates

Fitting a model● Full BiSSe model

○ Two speciation parameters

○ Two extinction parameters

○ Two transition rates

One of each parameter for each oviparity and viviparity

Fitting a model

Two speciation parameters

Speciation Rates

Freq

uenc

y

Speciation rate - viviparous taxa

Speciation rate - oviparous taxa

MLE estimate of parameter

Two extinction parametersFr

eque

ncy

Extinction Rates

Extinction rate - oviparous taxa

Extinction rate - viviparous taxa

Two transition parametersFr

eque

ncy

Transition Rates

Viviparity to oviparity transition rate

Oviparity to viviparity transition rate

Two transition parametersFr

eque

ncy

Transition Rates

Viviparity to oviparity transition rate

Oviparity to viviparity transition rate

1-2 reversals to oviparity

7-13 transitions to viviparity

23 transitions to viviparity

Support for viviparity as root state 7%

Support for oviparity as root state 93%

Strong support for oviparity as the root state of squamates

Oviparity

Viviparity

Support Values

Support for oviparity as root is strong

Conclusions

● Strongest support for ancestral oviparity● Viviparity associated with higher speciation● Much higher transition rate from oviparity to

viviparity than the reverse○ But, one or two reversals to oviparity are supported

with this method● Looking at a point estimate of topology can

be misleading

Thank you!● David Swofford● Dan Warren● Rich FitzJohn and Matt Pennell● Alex Pyron and Frank Burbrink

WrightAprilM