Date post: | 11-Jan-2016 |
Category: |
Documents |
Upload: | evan-lester |
View: | 214 times |
Download: | 0 times |
Modelling Phenotypic Evolution by
Stochastic Differential Equations
Combining statistical timeseries with fossil measurements.
Tore Schweder and Trond ReitanUniversity of Oslo
Jorijntje HenderiksUniversity of Uppsala
ICES 2010, Kent
1
Overview
1. Introduction:• Motivating example: Phenotypic evolution: irregular time series related by common
latent processes Coccolith data (microfossils) 205 data points in space (6 sites) and time (60 million years)
2. Stochastic differential equation vector processes• Ito representation and diagonalization• Tracking processes and hidden layers• Kalman filtering
3. Analysis of coccolith data• Results for original model• 723 different models – Bayesian model selection and inference
4. Second application: Phenotypic evolution on a phylogenetic tree• Primates - preliminary results
5. Conclusion
2
The size of a single cell algae (Coccolithus) is measured by the diameter of its fossilized coccoliths (calcite platelets). Want to model the evolution of a lineage found at six sites.•In continuous time and ”continuously”.•By tracking a changing fitness optimum.•Fitness might be influenced by observed (temperature) and unobserved processes.•Both fitness and underlying processes might be correlated across sites.
Irregular time series related by latent processes: evolution of body size in Coccolithus
Henderiks – Schweder - Reitan
19,899 coccolith measurements, 205 sediment samples (1< n < 400) of body size by site and time (0 to -60 my).
3
Our data – Coccolith size measurements
205 Sample mean log coccolith size (1 < n < 400) by time and site. 4
Individual samples
Bi-modality rather common.Speciation? Not studied here!
5
Evolution of size distribution
Fitness = expected number of reproducing offspring.
The population tracks the fitness curve (natural selection)The fitness curve moves about, the population follow.
With a known fitness, µ, the mean phenotype should be an Ornstein-Uhlenbeck process (Lande 1976).
With fitness as a process, µ(t),, we can make a tracking model: 6
)()()( tdWdttXtdX
)()()()( tdWdtttXtdX
The Ornstein-Uhlenbeck process
Attributes: • Normally distributed• Markovian• long-term level: • Standard deviation: s=/2α•α: pull• Time for the correlation to drop to 1/e: t =1/α
1.96 s
-1.96 st
The parameters (, t, s) can be estimated from the data. In this case: 1.99, t=1/α0.80Myr, s0.12. 7
)()()( tdWdttXtdX
One layer tracking another
Red process (t2=1/2=0.2,
s2=2) tracking black process (t1=1/1=2, s=1)
Auto-correlation of the upper (black) process, compared to a one-layered SDE model.
A slow-tracking-fast can always be re-scaled to a fast-tracking slow process. Impose identifying restriction: 1 ≥ 2
8
)()()(
)()()()(
22222
112111
tdWdttXtdX
tdWdttXtXtdX
Process layers - illustration
Layer 1 – local phenotypic expression
Externalseries
Fixed layer
Observations
T
Layer 2 – local fitness optimum
Layer 3 – hidden climate variations or primary optimum
Coccolith model
10
)),(( are nsObservatio
,
layer a inside correlatedpossibly white,is noise The
ackingperfect tr walk,random 0
layerth ' the tobelonging site of speed tracking theis 0
re)(temperatu process covariate observed)(
tEnvironmen )()(
optimum Fitness )()()()(
mean Population )()()(
size). (log at time siteat processth ' theof state )(
2,1,,
123
,,
,
3,3,3,
3,2,2,2,
2,1,1,1,
,
tssts
lsls
ls
sss
sssss
ssss
ls
tXNZ
XXuX
ls
tu
noisedttXtdX
noisedttutXtXtdX
noisedttXtXtdX
tsltX
Stochastic differential equation (SDE) vector processes
11
)())()(()(
matrixdiffusion ldimensiona
matrix pull ldimensiona :
matrix regression a process, level ldimensiona- :)()(
motion)(Brownian processener vector Wildimensiona- :)(
process state vector ldimensiona- :)(
tdWdtttAXtdX
qp
ppA
rpptut
qtW
ptX
matrix.r eigenvectoleft theis , ofmatrix eigenvalue diagonal a is
existing)(when !/)()(
)()()()()0()()(
:solution Ito
0
1
0 0
1
VAΛ
VeVntAetB
udWuBduuuBXtBtX
n
tnnAt
t t
Model variants for Coccolith evolution
The individual size of the algae Coccolithus has evolved over time in the world oceans. What can we say about this evolution?How fast do the populations track its fitness optimum?Are the fitness optima the same/correlated across oceans, do they vary in concert?Does fitness depend on global temperature? How fast does fitness vary over time?Are there unmeasured processes influencing fitness?
Model variations:• 1, 2 or 3 layers.• Inclusion of external timeseries• In a single layer:
Local or global parametersCorrelation between sites (inter-regional correlation)Deterministic response to the lower layerRandom walk (no tracking)
12
Likelihood: Kalman filter
13
The Ito solution
gives, together with measurement variances, what is needed to calculate the likelihood using the Kalman filter: and
t t
udWuBduuuBXtBtX0 0
1 )()()()()0()()(
iii mzzzE ),,|( 11 iii vzzzVar ),,|( 11
)()()( 21 ntXtXtX
1z 2z nz
Process
Observations
Need a linear, normal Markov chain with independent normal observations:
),,|()|( 1 iiiNin vmzfzf
Kalman smoothing (state estimation)ML fit of a tracking model with 3 layers
North AtlanticRed curve: expectancyBlack curve: realizationGreen curve: uncertainty
Snapshot:14
Inference
Exact Gaussian likelihood, multi-modal
Maximum likelihood by hill climbing from 50 starting points BIC for model comparison.
Bayesian: Wide but informative prior distributions respecting identifying restrictions MCMC (with parallel tempering) +
Importance sampling (for model likelihood) Bayes factor for model comparison and posterior probabilities Posterior weight of a property C from posterior model probabilities
15
' ''
)|(1
)|(1
)(
C CMC
CMC
Mdatafn
Mdatafn
CW
Model inference concerns
Concerns: Identification restriction: increasing tracking speed up the layers In total: 723 models when equivalent models are pruned out)
Enough data for model selection?• Data Simulated from the ML fit of the original model• Model selection over original model plus 25 likely suspects.
• correct number of layers generally found with the Bayesian approach. • BIC seems too stingy on the number of layers.
16
Results, original model3 layers, no regionality, no correlation
17
Bayesian inference on the 723 models
18
' ''
)|(1
)|(1
)(
C CMC
CMC
Mdatafn
Mdatafn
CW
95% credibility bands for the 5 most probable models
19
Summary
Best 5 models in good agreement. (together, 19.7% of summed integrated likelihood):
Three layers. Common expectancy in bottom
layer . No impact of exogenous
temperature series. Lowest layer: Inter-regional
correlations, 0.5. Site-specific pull.
Middle layer: Intermediate tracking.
Upper layer: Very fast tracking.
Middle layers: fitness optima
Top layer: population mean log size
20
Layered inference and inference uncertainty
21
Phenotypic evolution on a phylogenetic tree: Body size of primates
22
model. SDE thefrom sexpression parametric asmatrix
covariance andr mean vecto with ddistributenormally -multi be
will),0()( nsObservatio
.given tree for the equalities on theseCondition
. when )()(
. fromsplit when Time
: treeicPhylogenet
coccolithsfor as model trackingsame
optimumprimary t,environmen underlying)(
optimum fitness)(
mass) (log phenotypemean )(
fossils) 117 ts,measuremenextant (92 209,,1 tsMeasuremen
2,,
,,,
,
1,2,3,
3,
2,
1,
kkkiki
jiljli
ji
iii
i
i
i
NtXZ
TttXtX
jiT
XXX
tX
tX
tX
i
First results for some primates
23
Why linear SDE processes?
Parsimonious: Simplest way of having a stochastic continuous time process that can track something else.
Tractable: The likelihood, L() f(Data | ), can be analytically calculated by the Kalman filter or directly by the parameterized multi-normal model for the observations. ( = model parameter set)
Some justification from biology, see Lande (1976), Estes and Arnold (2007), Hansen (1997), Hansen et. al (2008).
Great flexibility, widely applicable...
24
Further comments
Many processes evolve in continuous rather than discrete time. Thinking and modelling might then be more natural in continuous time.
Tracking SDE processes with latent layers allow rather general correlation structure within and between time series. Inference is possible.
Continuous time models allow related processes to be observed at variable frequencies (high-frequency data analyzed along with low frequency data).
Endogeneity, regression structure, co-integration, non- stationarity, causal structure, seasonality …. are possible in SDE processes.
Extensions to non-linear SDE models… SDE processes might be driven by non-Gaussian instantaneous
stochasticity (jump processes).
25
Bibliography
Commenges D and Gégout-Petit A (2009), A general dynamical statistical model with causal interpretation. J.R. Statist. Soc. B, 71, 719-736
Lande R (1976), Natural Selection and Random Genetic Drift in Phenotypic Evolution, Evolution 30, 314-334
Hansen TF (1997), Stabilizing Selection and the Comparative Analysis of Adaptation, Evolution, 51-5, 1341-1351
Estes S, Arnold SJ (2007), Resolving the Paradox of Stasis: Models with Stabilizing Selection Explain Evolutionary Divergence on All Timescales, The American Naturalist, 169-2, 227-244
Hansen TF, Pienaar J, Orzack SH (2008), A Comparative Method for Studying Adaptation to a Randomly Evolving Environment, Evolution 62-8, 1965-1977
Schuss Z (1980). Theory and Applications of Stochastic Differential Equations. John Wiley and Sons, Inc., New York.
Schweder T (1970). Decomposable Markov Processes. J. Applied Prob. 7, 400–410
Source codes, examples files and supplementary information can be found at http://folk.uio.no/trondr/layered/.
26