t. bayes, phil. trans. roy. soc., 330 (1763). bayesian inference of phylogeny p(ti|s) probability of...

76
T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny T Ti p Ti S p Ti p Ti S p S Ti p ) ( ) ( ) ( ) ( ) ( Ι Ι Ι p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability or likelihood of the data S given tree Ti p(Ti) prior probability of Ti “The denominator sums the probabilities over all possible trees”

Upload: teodoro-moras

Post on 28-Jan-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

T. Bayes, Phil. Trans. Roy. Soc., 330 (1763).

Bayesian Inference of Phylogeny

T

TipTiSp

TipTiSpSTip

)()(

)()()(

Ι

ΙΙ

p(Ti|S) probability of the tree Ti given the sequence data Sp(S|Ti) probability or likelihood of the data S given tree Ti

p(Ti) prior probability of Ti

“The denominator sums the probabilities over all possible trees”

Page 2: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 3: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 4: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 5: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

ESTIMACION BAYESIANA• Inferencias están basadas en la

probabilidad de distribución posterior de un parámetro.

• La unión de las probabilidades de todos los parámetros son calculados.

• Las probabilidades están basadas en algún modelo (esperado a priori), luego de aprender algo de los datos.

Page 6: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

ESTIMACION BAYESIANA

Page 7: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

DADOS

Page 8: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

ESTIMACION BAYESIANA

• ¿Cuál es la probabilidad de tomar un dado trucado?

• Respuesta :1/10.

• Esta número representa la probabilidad a priori de tomar un dado sesgado.

Page 9: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

ESTIMACION BAYESIANA

Supongamos ahora que otra persona toma un par de dados de la caja y los tira.

Resultando:

¿Podemos creer que este resultado esta sesgado?

Dos aproximaciones: Maximum Likelihood e Inferencia Bayesiana.

Page 10: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

PROBABILIDADES

OBSERVACION NORMALES SESGADOS

Page 11: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

PR

PR

NORM

SESG

PROBABILIDADES

Page 12: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 13: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

ESTIMACION BAYESIANA

Page 14: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

Pr [Sesgados

INFERENCIA BAYESIANA

Page 15: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

ESTIMACION BAYESIANA

Page 16: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 17: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 18: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

11 44

Page 19: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

posterior

a priori

Page 20: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 21: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

Probabilidad a posteriori

Likelihood Probabilidad a priori

Σ de todas las probabilidades a posteriori

Integración de todas las posibles combinaciones de largo de ramas y modelos de sustitución nucleotídica.

Page 22: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 23: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

INFERIR UNA FILOGENIA

Page 24: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

POSIBLES FILOGENIAS

Page 25: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 26: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 27: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

Arboles equiprobables

Proporcional a observaciones: supuestos ej. alineamiento

Combinación: probabilidades a priori y Likelihood

Page 28: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 29: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 30: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

ALINEAMIENTO

Page 31: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 32: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 33: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 34: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 35: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 36: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 37: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 38: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 39: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 40: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 41: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 42: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 43: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

Estimación de las probabilidades a posteriori : ¿Cómo aproximarse?

• Calcular esta probabilidad implica: involucrar todos los árboles posibles….para cada árbol se debe integrar sobre todas las combinaciones de largo de rama y modelos de sustitución nucleotídica.

(IMPOSIBLE ANALÍTICAMENTE!!!) • Por necesidad la solución debe ser aproximada

• Método de Montecarlo

Page 44: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 45: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 46: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 47: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 48: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 49: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 50: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

Monte Carlo y cadenas Markovianas (MCMC)

• MCMC trabaja del siguiente modo:• a) Comienza una cadena markoviana con un

árbol ya sea 1) elegido al azar o 2) elegido por el investigador.

• b) Un nuevo árbol es propuesto….el proceso de cambio del arbol 1 al 2 debe satisfacer las siguientes condiciones:

1) El mecanismo debe ser estocástico; 2) cada arbol posible debe ser obtenido por aplicaciones repetidas del mismo mecanismo y 3) la cadena debe ser aperiodica.

Page 51: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 52: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

MARKOV CHAIN MONTE CARLO (MCMC)

Page 53: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

At each step in the chain a new tree is proposed by altering the At each step in the chain a new tree is proposed by altering the topology, or by changing branch lengths or the parameters of the topology, or by changing branch lengths or the parameters of the

model of sequence evolution.model of sequence evolution.

The Metropolis-Hastings algorithm is then used to accept or reject The Metropolis-Hastings algorithm is then used to accept or reject the new tree.the new tree.

Page 54: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 55: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 56: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 57: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 58: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

• Involucra correr algunas cadenas independientemente.

• La primera cadena que se cuenta (cold chain) el resto se denomina cadenas accesorias (heated chain).

• Saltos son intentados al azar entre dos cadenas distintas.

• Se necesita correr varios análisis independientes para confirmar convergencias.

METROPOLIS-COUPLED MARKOV CHAIN MONTE CARLO (MCMCMC o MC3)

Page 59: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 60: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 61: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 62: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 63: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 64: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

Resultado de esta búsqueda se obtiene un tercer término para la estimación de las probabilidades a posteriori (Proposal Ratio o Término de Hasting)

Page 65: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 66: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 67: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 68: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 69: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

INFERENCIA FILOGENÉTICA BAYESIANA

Phylogenetic tree

DNA Data

Evolutionary modelLikelihood

Prior probability

Posterior prob.

MCMC

Starting treeProposal

A sequence of Samples

inferencia

Approximate the distribution

Page 70: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability
Page 71: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

MrBayes: Bayesian Inference of Phylogeny

MrBayes is a program for Bayesian inference of phylogeny using Markov chain Monte Carlo methods. Avaialble for Mac, PC, and Unix.

Page 72: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

Métodos filogenéticos más usados

Data set

Algorithm

Algorithmicmethod

Optimization method

Distance matrix Character data

UPGMA

Neighbor-join

Fitch-Margolish

StatisticalSupported

Maximum Parsimony

MaximumLikelihood

Bayesian Methods

Search Strategy

Greedy search

Divide &Conquer

Stochastic search

DCM, HGT, Quartet

GA, SAMCMC

ExhaustiveBranch & Bound

Exact search

Stepwise additionGlobal arrangementStar decomposition

Page 73: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

Mapping characters onto phylogenies

Page 74: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

Mapping Uncertainty

parsimony ML

Bayesian

Page 75: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability

Phylogenetic and Mapping Uncertainty

Page 76: T. Bayes, Phil. Trans. Roy. Soc., 330 (1763). Bayesian Inference of Phylogeny p(Ti|S) probability of the tree Ti given the sequence data S p(S|Ti) probability