Ancestral reconstruction and phylogenetic signal of discrete traits in Mesquite

Primary Contact(s)	Created	Required Software
Rich Glor	28 March 2008	Mesquite
Example Datafile	Prerequisites
Exercise_2.zip	Finch exercise in Mesquite

In this exercise, we will expand on the lessons learned in the exercise on reconstruction of ancestral diets of Darwin's finches by using three datasets to further explore ancestral reconstruction. These datasets are larger and more complicated than the finch datasets considered previously (and also more realistic). In addition to investigating ancestral reconstruction (as we did with the finch example), we will also learn a basic test for phylogenetic signal (or phylogenetic inertia). This test reshuffles tips to ask whether the distribution of character states is determined, at least in part, by the phylogeny. Put another way, this test ask whether related species are more similar than expected by chance. Such tests of phylogenetic signal should be conducted in association with any phylogenetic comparative analysis. A lack of signal can be a blessing or a curse. In the realm of ancestral reconstruction, it tends to be more of a curse because we can’t reasonably use the phylogeny to reconstruct the evolution of traits that have no relationship to the tree.

With nearly 400 species distributed across the neotropics, ''Anolis'' lizards (anoles) are a classic example of adaptive radiation. The anole radiation is particularly diverse and abundant on Caribbean islands, where as many as 12 species may be found living sympatrically. Ecological specialization has long been considered key to explaining this diversity. For many years, anole biologists have been interested in knowing what traits characterized the ancestral anole that spawned the impressive Caribbean radiation.

1. Open the file titled “GA_Anolis_Characters.txt” in Mesquite (NOTE: Mesquite will permit you to have multiple projects open simultaneously, but it might be best to close the finch datasets prior to proceeding with subsequent exercises to avoid confusion). This file contains data for 85 species of Anolis lizards found on the four large Greater Antillean islands of the northern Caribbean. After opening this file, you should see a Character Matrix with three columns. The first two columns represent two distinct types of ecological specialization. The first represents the categories of microhabitat specialists that are commonly referred to as “ecomorphs”: 0=grass-bush, 1=trunk-ground, 2=trunk, 3=trunk-crown, 4=crown-giant, 5=twig. The second column represents macrohabitat specialization: 0=lowland mesic forest, 1=lowland xeric forest, 2=montane forest. The final column identifies the island that each species is found on: 0=Cuba, 1=Hispaniola, 2=Jamaica, 3=Puerto Rico.

2. As in the previous exercise, we want to include a linked tree file by going to the “File” tab, selecting “Include file...”, and then selecting “GA_Anolis_Tree.nex”. To view this tree you can either go to the “Taxa&Trees” tab and open a new window or simply click on the Tree block in the Projects and Files window.

3. We now want to reconstruct ancestral states. To do this with parsimony, just follow steps 7-8 from the previous Mesquite exercise. What does parsimony tell us about the microhabitat of the ancestral anole? If all has gone according to plans, you should have found that the twig ecomorph is the most likely ancestral state. This is a very curious conclusion as the twig anoles are among the most narrowly specialized of all anoles. Let’s see what maximum likelihood has to say...

4. Follow step 9 from the previous Mesquite exercise to reconstruct ancestral states using maximum likelihood. After doing this, you will see that likelihood reconstructs considerable uncertainty in the state of the ancestor to the Greater Antillean radiation. Thus, the surprising conclusion that the ancestor of the anole radiation is a twig anole is no longer supported.

5. To view ancestral reconstructions for macrohabitat, follow the steps above and simply click the blue arrow in the “Trace Character” palette to view the second character in the Character Matrix. What do parsimony and likelihood tell you about the most likely macrohabitat of the radiation’s ancestor? Although both methods suggest a lowland mesic ancestor, the states appear to somewhat randomly distributed across the tips. Could it be that we are reconstructing a mesic ancestor simply because this is the most common state rather than due to some information present in the phylogeny itself. To address this question, we’ll test the amount of phylogenetic signal present in this, and other, characters.

6. Go to the “Analysis” tab and select “New Bar & Line Chart For:Trees”.

7. Select: Randomly modify current tree:Reshuffle terminal taxa:Steps in character:Stored Characters:Current Parsimony Model.

8. After the following series, you should be prompted with an alert, click “OK” and enter the number of times you want to reshuffle the data (the default of 100 should be fine for our purposes) and click “OK”.

9. To identify outliers go to the “Chart” tab and select “Analysis:Percentiles”. Select the desired threshold (the default of 0.05 is suitable here) and click “OK”. Values that fall outside of these bounds contain significantly more or less structure than expected by chance. Values that have fewer changes than the red line on the left are characterized by significant phylogenetic signal. Based on the number of changes in microhabitat/ecomorph inferred in the most parsimonious reconstruction of the original data, does microhabitat exhibit significant phylogenetic signal?

10. To investigate signal in macrohabitat go to the “Chart” tab and select “Next Character”. Does macrohabitat exhibit significant phylogenetic signal? Given the results of this analysis what are we able to say about signal? What does a lack of phylogenetic signal mean for our ability to infer the most likely ancestral state?

A group of plants from western North American known as columbines represent another classic example of adaptive radiation. Much like anoles, columbines (genus ''Aquilegia'') have diverged along several distinct ecological axes. Two important axes of divergence involve specialization to different pollinators and habitats. Here we will categorical data for these two axes to ask similar questions to those considered above in Anolis. For more details on pollinator evolution, check out Justen Whittal's rad paper on this topic (Whittall_etal_2007.pdf).

Figure 1 (Figure 3 from Whittall et al. 2007) - Reconstructed history of pollinator syndromes in Columbine flowers. Blue indicates bee pollinated, red indicates hummingbird pollinated, and yellow indicates hawkmoth pollinated.

Part 1: Open project in Mesquite & reconstruct ancestral states

1. Open the file “Aquilegia.nex” in Mesquite. This file contains data on two ecological traits for 31 species of columbines. After opening this file, you should see a Character Matrix with two columns labeled ‘pollinator’ and ‘habitat’. In the pollinator column: 0=bee, 1=hummingbird, and 2=hawkmoth. This file should already include the tree file that you will be using for these analyses.

2. By now you should be familiar enough with ancestral reconstruction and the simple reshuffling test of phylogenetic signal to experiment on your own (or you can just refer back to the instructions for the previous anole example).

3. What have you learned about the ancestor of the columbine radiation? Do pollinator syndrome and habitat exhibit phylogenetic signal?

Whittall, J. B., and S. A. Hodges. 2007. Pollinator shifts drive increasingly long nectar spurs in columbine flowers. Nature 447:706-709. Whittall_etal_2007.pdf

Comments:

Note: You must be logged in to add comments

2009-04-09 02:34:44   i would want to learn more —137.158.152.206

2010-03-26 11:20:36   When I do the maximum likelihood ancestral reconstruction in Mesquite of pollinator character in Aquilegia there are nodes that differ significantly from Whittall estimation. What could be the reason? —132.248.207.161