Non independence and convergence mislead morphological phylogenetics of phyllostomids
1. Lying through your teeth: non independence and
convergence mislead morphological phylogenetics
of phyllostomids
Liliana M. Dávalos —Stony Brook University
Paul M. Velazco —American Museum of Natural History
Omar M. Warsi —Stony Brook University
Peter D. Smits —University of Chicago
Nancy B. Simmons —American Museum of Natural History
3. Genome not always
available
•Majority of species are
extinct
•Fossils are all that
remain
•Phylogenies must use
morphology
•How?
Morgan Czaplewski 2012 Evolutionary
History of Bats
4. Total evidence Conditional combination
Hermsen Hendricks 2008 Ann Missouri Springer et al. 2007 Syst Biol
Bot Gard
5. Assumptions of
phylogeny
•Homology: character
changes reflect
common descent
•IID: Independent and
Identically Distributed
7. Saturation is not
everything
•If rates of evolution are
high, then signal
erased over time
•Results in
unresolved
phylogeny
•Other signal must
emerge to resolve
phylogeny
• First position
• Second position
o Third position
Dávalos Perkins 2008 Genomics
10. 15.6
10.4
5.2
8.5
6.3
4.2
0
Frequency (percent)
Morphological
0
Molecular
2.1
10 A
Dissimilarity between characters
Relative density between morphological characters
0.0 0.2 0.4 0.6 0.8 1.0
8
6
4
2
0
B
Figure 3
Signal is amplified by
repetition
•Measured dissimilarity
between pairs of
characters
•High dissimilarity
among molecular
characters
•Despite protein-coding
loci
•This is not the case for
dental characters
Dávalos et al. Accepted Syst Biol
11. How similarity arises
•Hypotheses:
•Occlusion
•Negative selection
•Ecological
convergence
•Positive selection
•Development
•Paths of least
resistance
1!
2!
1! 2!
13. Data models don’t
match
•Less is more when
collecting certain kinds of
characters
•Dental data violate key
assumptions of
phylogenetic models
•Saturation, convergence,
and non-independence
•= model failure
•New data models needed
Czaplewski et al. 2003 Caldasia
14. No
conflict
Morphological
phylogenies
Statistical scaffold
RQÀLFWLQJ
characters
Starting point
Dependency
Morphological
Rates
Data matrix
Statistic from
data matrix
Phylogeny
Statistic from
phylogeny
Phylogenetic
analysis
Other
analysis
Comparison
Endpoint
Alternative
combined
phylogenies
)LJXUH
Pairwise
dissimilarity Bayesian
State:step
Simulations
Distribution
morphological
dissimilarity
Pairwise Bayesian
dissimilarity
Distribution
molecular
dissimilarity
Agree?
Yes
No
Null distribution
difference
per-character
likelihoods
Tree with
Tree
without
Backbone
constraints
Bayesian
Branch lengths
Maximum
likelihood
Constrained
phylogenies
Molecular
phylogenies
Combined
phylogenies
Combined
matrix without
FRQÀLFWLQJ
characters
Variable sites
Relative
distribution
RQÀLFWLQJ
nodes
Relative
rates of
evolution
RQÀLFWLQJ
nodes
Branch
lengths
RQÀLFWLQJ
morphological
characters
Statistical
scaffold
Morphological
saturation
plot
Morphological
Combined
Molecular
Analytical innovations
•Statistical scaffolds =
condition
morphological
resolution on
molecular posterior
•Morphological
parametric bootstraps
= uncover significantly
conflicting characters
15. Thanks!
• Funding
• NSF—DEB
• CIDER—SBU
• Speciation diversification: A.
Cirranello, E. Dumont, A. Russell,
N. Gerardo, A. Wilson
• Dávalos Lab
• Phylogenetics: B. Baird, S.
DelSerra, A. Goldberg, O.
Warsi, L. Yohe