1. An Introduction to Phylogenetic Methods
Part one
Dr Laura Emery

2. Objectives After this tutorial you should be able to…
Discuss a range of methods for phylogenetic inference, their advantages, assumptions and limitations
Implement some phylogenetic methods using publicly available software
Appreciate some approaches for assessing branch support and selecting an appropriate substitution model
We don't have enough time to go into any of these methods in great depth, and if you want to fully understand how they work, then you will need to go away and do some extra reading. But the point of this tutorial is to make you aware of some of the most commonly available methods, their assumptions and limitations. And also to direct you to information that you can read up on if you want further help

3. Outline Alignment for phylogenetics
Phylogenetics: The general approach
Phylogenetic Methods (1 – simple methods)
Assessing Branch Support
BREAK
Substitution Models
Phylogenetic Methods (2 - statistical inference)
Deciding which model to use (hypothesis testing)
Software
Distance-based methods (UPGMA , NJ)
Maximum Parsimony
Substitution models (JC, K2P, F81, HKY85, GTR + gamma)
Other substitution models
Maximum likelihood
Bayesian inference

4. Alignment for phylogenetics
Phylogenetic analyses are typically applied to alignments of sequence data
Occasionally other data such as morphological traits are used (e.g. when no sequence data is available)
Alignments must contain homologous sequences
We assume that sites in the same column in an alignment are homologous

5. Alignment for phylogenetics
Benjamin Redelings

6. Columns in alignments should be homologous
Benjamin Redelings

7. Phylogenetics: The general approach
We want to find the tree that best explains our aligned sequences
We need to be able to define “best explains”
we need a model of sequence evolution
we need a criterion (or set of criteria) to use to choose between alternative trees
then evaluate all possible trees
(NB: if N=20, then 2 x 1020 possible unrooted trees!)
or take a short cut
Paul Sharp

8. There is only one true tree
The true tree refers to what actually happened in the evolutionary past
All methods attempt to reconstruct the true phylogeny
Even the best method may not give you the true tree
Use the thought experiment of thinking of your ma'am and then your mothers ma'am and so on into the evolutionary past.
This is something we must be mindful of when implementing or assessing any phylogenetic inference of what we think might have happened in the past based upon the data and our beliefs about how sequences evolved
No model is a true description of the biological complexity that has happened. They are simplifications and approximations which are useful as a tool to help us figure out what has happened. And for some taxa this may be something which we never resolve.

9. Methodological approaches
Distance matrix methods (pre-computed distances)
UPGMA assumes perfect molecular clock Sokal & Michener (1958)
Minimum evolution (e.g. Neighbor-joining, NJ) Saitou & Nei (1987)
Maximum parsimony Fitch (1971)
Minimises number of mutational steps
Maximum likelihood, ML
Evaluates statistical likelihood of alternative trees, based on an explicit model of substitution
Bayesian methods
Like ML but can incorporate prior knowledge
We will examine these methods in this order
We will examine the UPGMA method in detail because it is simple and gives you an idea of some of the pitfalls of other methods, although it is not commonly used any more
will examine overview of the methods, although there is not time for us to look at exactly how they work in detail, but we will consider their assumptions advantages and limitations

10. What is a distance matrix?
A table that indicates the number of substitutions between pairs of sequences

11. Distance Matrix Methods
Take alignment
compute genetic distance for all pairwise combinations
put it into a matrix
cluster together those sequences which are most similar
recalculate the matrix using a single distance for the sequences you've clustered together
continue doing this until the trees finished
Andrew Rambaut

12. UPGMA Method
Identify the pair of most closely related taxa according to the pairwise-genetic distance matrix
Cluster these together
Figures Andrew Rambaut

13. UPGMA Method
Recalculate distance matrix (calculate the distances from the new cluster to every other sequence)
Take the average of both distances
E.g. distance[spinach, monkey/human] :
= (distance[spinach, human] + distance[spinach, monkey]) / 2
= ( )/2 = 88.55
Figures Andrew Rambaut

14. UPGMA Method Repeat the procedure until the tree is finished
Produces rooted phylogeny
distance between (spi,ric) and mos(mon,hum) is 108.7
Andrew Rambaut

15. UPGMA Method Assumptions: Advantages: Disadvantages:
Strict molecular clock
Ultrametric distance data
Advantages:
Fast and simple
Disadvantages:
Data are almost never ultrametric
Usage: Almost never used
Ultra-metric = genetic distance data exactly proportional to time

16. Neighbour Joining Method
An improvement over the UPGMA: does not require data to be ultrametric
Identifies the topology that gives the least total branch length at each step
after we have joined two species in a cluster we have to
compute the distances from every other sequence to the new
cluster.
We do this with a simple average of distances:
distance[spinach, monkey/human]
= (distance[spinach, human] + distance[spinach, monkey]
- distance[monkey, human]) / 2
Figures Olivier Gascuel

17. Neighbour Joining Method
Advantages:
allows the use of an explicit model of evolution
fast and simple
able to deal with thousands of taxa
Disadvantages:
only produces one tree
reduces all sequence information into a single distance value
dependant on the evolutionary model used
Usage: commonly used due to being widely available in many software packages

18. Methodological approaches
Distance matrix methods (pre-computed distances)
UPGMA assumes perfect molecular clock Sokal & Michener (1958)
Minimum evolution (e.g. Neighbor-joining, NJ) Saitou & Nei (1987)
Maximum parsimony Fitch (1971)
Minimises number of mutational steps
Maximum likelihood, ML
Evaluates statistical likelihood of alternative trees, based on an explicit model of substitution
Bayesian methods
Like ML but can incorporate prior knowledge
We will examine these methods in this order
We will examine the UPGMA method in detail because it is simple and gives you an idea of some of the pitfalls of other methods, although it is not commonly used any more
will examine overview of the methods, although there is not time for us to look at exactly how they work in detail, but we will consider their assumptions advantages and limitations

19. Maximum Parsimony
The most parsimonious tree is the tree requiring the smallest number of substitutions to explain the sequences * T A C
length = 2
length = 3 ? C A MP
(unrooted) T C A *
length = 3
Add to this the four diagrams that Sarah said she would do. Animate them so that initially there is one without the ancestral states, so that you can explain how we don't know what they were. Then pencils in is a possibility (not the most parsimonious tree). Then get the other 3 to appear, and finally the number of substitutions to appear
In practice, all possible trees need to be evaluated

20. Maximum Parsimony Assumptions: Advantages: Disadvantages
Multiple substitutions rare
Advantages:
fast
Disadvantages
not consistent with most models of evolution
can result in multiple optimal trees
Usage: still used with morphological data
Figures Andrew Rambaut

21. The problem of multiple substitutionsA * *
hidden mutations G A A T
More likely to have occurred between distantly related species
> We need an explicit model of evolution to account for these (to be covered in part two)
The further back in time you go, the more likely it is that hidden mutations have occurred, that you cannot see in the sequence data. You will never observe 100% sequence difference, because there are only four bases, and so you expect a maximum sequence difference of 75% (even with random sequences)

22. Methodological approaches
Distance matrix methods (pre-computed distances)
UPGMA assumes perfect molecular clock Sokal & Michener (1958)
Minimum evolution (e.g. Neighbor-joining, NJ) Saitou & Nei (1987)
Maximum parsimony Fitch (1971)
Minimises number of mutational steps
Maximum likelihood, ML
Evaluates statistical likelihood of alternative trees, based on an explicit model of substitution
Bayesian methods
Like ML but can incorporate prior knowledge
How well supported are my branches?
We will examine these methods in this order
We will examine the UPGMA method in detail because it is simple and gives you an idea of some of the pitfalls of other methods, although it is not commonly used any more
will examine overview of the methods, although there is not time for us to look at exactly how they work in detail, but we will consider their assumptions advantages and limitations

23. How well supported are my branches?
A tree is a collection of hypotheses
so we assess our confidence in each
of its parts or branches independently
There are three main approaches:
Bootstraps
Bayesian methods
Approximate likelihood ratio test (aLRT) methods
0.93
0.81
0.99 85 63
100
We will look at the straps in more detail
probabilistic

24. Bootstrapping
2. Resample columns with replacement to create many dummy alignments
1. Take your alignment, and consider each column separately
repeat lots
repeat lots
3. Use these to draw many trees and count up the occurrences of each branch among these trees
Figures Andrew Rambaut
Felsenstein, J Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39:

25. Issues with bootstrapping
Sites may not evolve independently
P values are biased (too conservative)
Calculating bootstraps for many branches results in multiple testing
Bootstrapping does not correct biases in phylogeny methods
Nevertheless they perform surprisingly well

26. Outline Alignment for phylogenetics
Phylogenetics: The general approach
Phylogenetic Methods (1 – simple methods)
Assessing Branch Support
BREAK
Substitution Models
Phylogenetic Methods (2 - statistical inference)
Deciding which model to use (hypothesis testing)
Software
Distance-based methods (UPGMA , NJ)
Maximum Parsimony
Substitution models (JC, K2P, F81, HKY85, GTR + gamma)
Other substitution models
Maximum likelihood
Bayesian inference

27. Now it's your turn… Open your course manuals and begin Tutorial 1
Also available to download from: bioinformatica-2013
You will require the alignment file 5SrRNA.txt
There are answers available online but it is much better to ask for help!

28. Thank you!
Facebook: EMBLEBI