inferring phylogenies

How do we make phylogenetic trees?

• Parsimony

• Maximum likelihood

• Bayesian phylogenetics

• Coalescence

• Supertree

Parismony

Characters and character states

What characters to use in constructing a tree?

• Apomorphies – derived characters changed from reference state

• Synapomorphies – shared, derived characters

• Autapomorphies – derived characters found only in 1 taxon

• Symplesiomorphies – shared, ancestral characters

Have to know what is ancestral vs. derived

• Outgroup – a taxa we know a priori that is outside, but fairly closely related to the

• Ingroup – the taxa we want to know the relationships among

Ways to define a good outgroup

• Previous phylogenetic knowledge (most common)

• Embryology – rare and problematic

• Fossil record – rare and problematic

In general, synapomorphies define

monophyletic clades in a tree

Not really useful, since all in-group taxa share it

Not really useful, Only 1 taxa has it

Steps in inferring a tree via parsimony

1) Identify outgroup

2) Identify useful characters (synapomorphies)

3) Map changes in character state onto each possible tree

4) The one with the fewest total # of changes is most parsimonious

what are some problems with parsimony?

Mutated independently but look same so parsimony makes them seem related divergent evolution

Environment makes them look similar not genetically similar tho

If two trees have same amount of changes can’t tell which parsimony is better

When is parsimony useful

• When traits are complex (e.g., some morphology)

• When reversals and homoplasy are rare/unlikely

• When there are lots of characters and character states (prevents ties)

• Usually need at least as many characters as taxa

• Not when using DNA or protein data

Using DNA/Protein data to build trees

• Need to first generate an alignment (can learn the basics of this in a computational

biology class) to make sure each position is homologous

• Many characters, only 4 character states (reversals and Homoplasy are rampant)

• Should use a model of base or amino acid substitutions when dealing with with proteins

Maximum likelihood as an alternative method

• Computationally intensive method where probabilities are estimated to generate a

likelihood score for all parameters of interest, which could include:

• Tree topology

• Length of each branch in the tree

• Rates of change between different bases

• Proportions of different types of sites in the alignment

• Substitution model

• Etc

• Try to maximize likelihood score by Iteratively sampling “tree space”

Nodal support Bootstrapping

Break down alignment into different subsections of the data

Well supported closer to 100

Not well supported smaller number and collapsed into polytomy

Genomic phylogenetics

• Maximum likelihood is one of several methods

• Applying phylogenetic inference to whole genome data is an ongoing area of development

• Which genes to use and why different regions give different signals is an important area of phylogenetics

Reasons for phylogenetic incongruence

• Noise due to limited data/sampling, wrong parameters

• Introgression/hybridization

• Incomplete lineage sorting

• “Hard” polytomies

Why make trees?

• They can be used to both generate and test hypotheses

• A phylogeny itself is a hypothesis – a hypothetical scenario describing

common ancestors and descendants in a group

• This hypothesis can change when more data is added

Use phylogenies to estimate rates of,evolution