The goal of this tutorial is to illustrate how to create an package for BEAST 2. You will create an package that adds a new nucleotide substitution (F84) model to BEAST 2. The tutorial assumes that the reader is using MacOS 10.8.5, but the tutorial should be reasonably portable. This tutorial was created by the “Add-On Tutorial and SDK” breakout group at the third NESCent BEAST meeting, Dec. 5-8, 2011, and revised at the fourth NESCent BEAST meeting, Dec. 3-6, 2013. Please contact Paul Lewis, Peter Beerli, or Remco Bouckaert with comments/suggestions.
We will use the Eclipse IDE to manage the new package project, so the first step is to install Eclipse if you do not already have it on your system. Go to http://www.eclipse.org/downloads/and download “Eclipse IDE for Java Developers” for your system. For our MacOS 10.8.5 system, we unzipped the downloaded eclipse-jee-kepler-SR1-macosx-cocoa-x86_64.tar.gz file by double-clicking it, then moved the unpacked folder (named eclipse) to our Applications folder. You may wish to drag Eclipse.app to your dock to make it easy to open. On opening Eclipse for the first time, you will be asked to select a workspace. For the purpose of this tutorial, we will assume you chose this directory:
Downloading the BEAST 2 source
Now, open Terminal app (in the MacOS Utilities folder) and navigate to the Eclipse workspace directory by typing this command:
Assuming git is installed, you should be able to download the BEAST 2 source code using this command:
Creating a BEAST 2 project
Back in Eclipse, use File > New > Project… > Java project to create a new Java project named “beast2”. You should be able to leave everything in the New Java Project dialog box at its default value. Once you press the Finish button, Eclipse will proceed to compile “beast2”.
Creating a project for “MyAddOn”
Now you need to create an Eclipse project for your package and make it depend on the “beast2” project.
o Use File > New > Java project to create a new Java project named “MyPackage”. Again leave everything in the New Java Project dialog box at its default setting and press the Finishbutton. Now you should see two projects in the Project Explorer pane: “beast2” and “MyPackage”. If you see Package Explorer rather than Project Explorer, use Window > Show View > Project Explorer to bring up the Project Explorer view instead.
o Right-click (or Control-click if you do not have a mouse with two buttons) “MyPackage” and choose Properties from the speed menu that appears
o Choose Java Build Path, click the Projects tab, click the Add… button, then check “beast2” and press the OK button. Now press the OK button in the Properties for MyPackage dialog box to close it.
Creating a package for “MyPackage”
With “MyPackage” selected in the Project Explorer, use File > New > Package to create a package. The Source folder for the package should be automatically filled in correctly (“MyPackage/src”), and the name should be set to “beast.evolution.substitutionmodel”. If Eclipse does not allow you to press the Finish button, cancel the dialog box, locate “beast.evolution.substitutionmodel” in the “beast2” project, copy it, then try again to create the package, this time pasting in “beast.evolution.substitutionmodel”.
Creating an F84 class
Now right-click (or Control-click) on the “beast.evolution.substitutionmodel” package under “MyPackage” and choose New > Class from the resulting popup menu. Give the new class the name “F84” and press Finish. You should now see the F84.java file open in the editor, and it should contain the following text: package beast.evolution.substitutionmodel;
Add “extends SubstitutionModel.Base” following the class name to produce the following:
At this point you should see an icon beside the class declaration. The icon consists of a light bulb with a red square containing a white letter X. Clicking this should bring up a popup menu that allows you to Add unimplemented methods. Your F84.java file should now look like this:
Note: depending on your particular version of Eclipse (or some other as-yet-unidentified factor), the methods that appear may not be in the order listed above. These are methods that must be implemented because they are abstract methods in the
SubstitutionModel.Base base class.
Adding the k parameter
The F84 model comprises nucleotide frequency parameters as well as a parameter k that determines the number of extra transition-type substitutions (above and beyond those generated by the general substitution process). If k equals zero, then the F84 model collapses to the F81 model, which allows nucleotide composition to vary but does not allow transition/transversion bias. As k becomes larger, transitions become more common. The nucleotide frequencies are supplied by the base class, but the k parameter is unique to the F84 model and thus must be defined by our package class. To do this, add the following line to the class definition just after public class:
kF84 parameter will require you to insert some declarations at the top of the file. Eclipse helps you do this:
o First, save the file
o Click the light-bulb icon to get a popup menu, and double-click the first item in the popup menu: Import ‘Input’ (beast.core).
o Continue clicking the light-bulb icon until you have all three of these declarations inserted:
Adding the initAndValidate method
We must also override the
Implementing the canHandleDataType method
The F84 model is a substitution model applicable only to nucleotide sequences, so replace the body of the automatically-created stub with the following:
The method now returns true if dataType is an instance of the Nucleotide class, and otherwise throws an exception with an informative message. Adding this code requires you to insert the following declaration at the top of the file:
Implementing the getEigenDecomposition method
For now, leave this method unimplemented. We will return to it later after testing the basic implementation.
Implementing the getTransitionProbabilities method
The getTransitionProbabilities method is expected to fill a 16 element vector with transition probabilities in the order AA, AC, AG, AT, CA, CC, CG, CT, GA, GC, GG, GT, TA, TC, TG, TT, where the first nucleotide in each pair is the “from” state and the second in each pair is the “to” state. Each transition probability depends on the edge length, which is measured in terms of the expected number of subsitutions. The expected number of substitutions is the product of rate (provided by
fRate) and time (the difference between
fStartTime). Note that time is viewed from the present looking toward the past, so
fStartTime - fEndTime is the correct way to get the time difference. Fill in the
getTransitionProbabilitiesmethod as follows:
Adding a description and citation
We are almost ready to try out the new package, but BEAST 2 forces us to provide a brief description of the F84 class. To do this, add the following line just before
public class F84 extends SubstitutionModel.Base:
(Ignore the light bulb icon for the moment.) We can also add citations, which are recommended to users of the package as papers they should cite. Add the following line just after the @Description:
There should be no line breaks in either the
@Citation lines: if lines are broken, you should fix these manually so that each of these declarations occupies just one line (a long one in the case of
You should now see two more light-bulb-redX icons, one beside the
@Description line and another beside the
@Citation line. Clicking these icons may result in a message like this: “Description cannot be resolved to a type” and no possible actions are suggested. If this happens (and we’re still not sure why it happens), hover your mouse over the words
Citation, in turn, and this should give you a ‘quick fix’ option to Import ‘Description’ (beast.core) and Import ‘Citation’ (beast.core), respectively. After choosing these ‘quick fixes’, you should see this import declarations at the top of your file:
Of course, you can always just type in the declarations directly rather than using Eclipse’ ‘quick fixes’
Debugging the new package inside Eclipse
Create an example XML file
Because the F84 model is so similar to the HKY model, we can copy the textHKY.xml file and replace the HKY-specific parts with F84 counterparts to obtain an XML file that can be used to test our F84 package. The following instructions show how to do this.
Create an examples folder
Right-click “MyAddOn” in the Project Explorer pane of Eclipse to obtain a popup menu, then choose New > Folder. Name the new folder “examples”.
Copy and rename the testHKY.xml file
In the Project Explorer pane of Eclipse, press the disclosure triangle beside the “beast2” project to expand, then expand the “examples” folder within “beast2” and copy the testHKY.xmlfile. Now paste it into the new “examples” folder inside the “MyAddOn” project. To rename the testHKY.xml file to testF84.xml, right-click it to bring up the popup menu, then select Rename….
Replace HKY/kappa with F84/kF84 throughout testF84.xml
The testF84.xml file still has several references to the HKY model. These need to be changed to be references to the new F84 class. To edit the file in Eclipse, double-click its name in the Project Explorer pane. This will likely drop you into Design view; to switch to the more useful Source view, click on the Source tab at the bottom left corner of the editor pane. Use the Edit > Find/Replace… menu command to bring up the Find/Replace dialog and (case-sensitively) replace all instances of “HKY with “F84” and all instances of “kappa” with “kF84”, then File > Save the file.
Create a debug configuration in Eclipse
To create a new debug configuration, follow these steps.
o First, open the “beast2” project in the Project Explorer pane, then expand “src”, then expand “beast.app.beastapp” and, finally, select “BeastMain.java”
o Use Run > Debug As… > Java Application. This will bring up a BEAST dialog box, but by the time this dialog box appears, a new debug configuration has been created, so you can immediately press the Quit button to dismiss the dialog box.
o Now that a debug configuration exists, you must modify it to debug your new package rather than BEAST 2.
o Use Run > Debug Configurations… to bring up the Debug Configurations dialog
o Change the Project from “beast2” to “MyAddOn” by clicking on the Browse... button, then choosing “MyAddOn”
o Click on the Arguments tab and type “examples/testF84.xml” in the Program arguments field to specify that the testF84.xml file should be run when debugging
o While still on the Arguments tab, enter
-Djava.only=true in the VM arguments field. This is really only necessary if you have the BeagleLib GPU library installed. Setting
true means that BeagleLib will not be used even if it is available. Later, we will implement the
getEigenDecomposition function in the F84.java file and, after that, BeagleLib can be used in analyses involving the new F84 model.
Running in debug mode
To run in debug mode, go the the main Eclipse Run menu, choose Debug configurations…, select “BeastMCMC” under Java Application, then press the Debug button to start. After running this configuration once, you can simply choose Run > Debug History > BeastMCMC to start debugging (and you will find a convenient button on the main toolbar that provides a shortcut to a “BeastMCMC” debug session as well).
If the debug session starts successfully, you should see output similar to that below:
Adding the F84 model to BEAUti
It is convenient to incorporate new packages into BEAUti so that users of your package are not required to hand-craft an XML file. Incorporating an package into BEAUti requires adding a template to your package project. The template tells BEAUti what information to collect for your package.
Create a template
First, create a new folder called templates under “MyAddOn” inside the Project Explorer pane in Eclipse. The easiest way to do this is to right-click (or Control-click) the “MyAddOn” project, then choose New > Folder from the resulting popup menu.
Now create a text file within the new templates folder named F84-beauti-template.xml. The easiest way to do this is to right-click (or Control-click) the templates folder within the “MyAddOn” project, then choose New > File from the resulting popup menu, giving the new file the name “F84-beauti-template.xml”. Copy the text below into the new F84-beauti-template.xml file and save it.
To create a debug configuration that runs BEAUti, choose Run > Debug Configurations… from the main Eclipse menu. In the Debug Configurations dialog, select Java Application and click the New launch configuration button. In the Name field at the top of the dialog box, type “BEAUti”. In the Project field of the Main tab, type (or use the Browse… button to select) “MyAddOn”. For the Main class, type “beast.app.beauti.Beauti”.
Now you should be able to click the Debug button to run BEAUti. Try using the following steps to generate a BEAST 2 XML file:
o Click the + button at the lower left, then navigate to the beast2/examples/nexus folder and select the anolis.nex file. (You can avoid doing this step if you specify “-nex ../beast2/examples/nexus/anolis.nex” in the “Program arguments” field of the Arguments tab in the BEAUti debug configuration.)
o Click the Site Model tab and select the new F84 model in the Subst Model drop down list
o Click the MCMC tab, and set Chain Length to 1000000 (the default is 10 million, which will take longer than you probably want to wait for a test example)
o Expand the tracelog logger and make sure Log Every is set to 1000
o Use File > Save As… to save the XML file under the name f84example.xml in the MyAddOn/examples folder
Run the f84example.xml file
The goal of this section is to create a debug configuration that is essentially a copy of the first one we created (in the section entitled “Create a debug configuration in Eclipse”) with the exception that the file that is run is f84example.xml instead of testF84.xml. The easiest way to create the new configuration is to simply duplicate the existing one. Here are instructions for accomplishing that:
o Use the menu command Run > Debug Configurations… to open the Debug Configurations dialog box
o Under Java Application, select the “BeastMCMC” configuration
o Right-click (Control-click on a Mac) and choose Duplicate from the popup menu
o Select the new configuration, which for us was automatically named “BeastMCMC (1)”
o Click the Arguments tab and change the Program arguments field to examples/f84example.xml
o If necessary, move the f84example.xml file inside the MyAddOn/examples directory so that its location matches the specification you just provided in the Program arguments field
You should now be able to click the Debug button to start an analysis of the f84example.xml file.
The maximum likelihood estimates on a neighbor-joining tree evaluated using PAUP* were:
The corresponding posterior means we obtained from running the F84 package in BEAST 2 were as follows:
The Bayesian estimate is lower than the maximum likelihood estimate due to the fact that we used the default prior on k, which is LogNormal with mean and standard deviation (on the log scale) of 1 and 1.25, respectively. Half the mass of this prior lies below 0.46, which pulls the posterior mean down slightly.
Preparing your package for release
In order to publish your package, you will need to compile it outside of Eclipse and build a jar file that you can distribute. Here is an overview of the steps involved:
o compile your package using ant to create a zip file
o get your package published on the BEAST 2 web site so that others can install it over the internet
Compile beast2 using ant
Navigate (in a Terminal window) to the beast2 project directory. If you have been following this tutorial faithfully, the folder should be $HOME/Documents/workspace/beast2. Type ant at the command line. (If ant cannot be found, you may need to install it from the Apache Ant website.) This will build BEAST 2 using the build.xml file supplied with BEAST 2. It is necessary to build BEAST 2 again (even though we’ve already built it in Eclipse) because Eclipse places its object files and intermediate build products in a different directory than does ant.
Compile MyAddOn using ant
The next step is to create a build.xml file within the $HOME/Documents/workspace/MyAddOn folder that tells ant how to build MyAddOn. Create a new text file (using Eclipse or any way you like) just inside the MyAddOn directory. Name the new file build.xml and enter the text below into the new file and save it.
You can now compile your package by changing directory to the MyAddOn folder and typing
Move the package to a place where BEAUti can find it
Once ant has finished compiling the package, you should have a beast2/build/dist/F84.addon.zip file inside your Eclipse workspace directory. For testing purposes, this file should be copied to $HOME/.beast2/F84 and unzipped. If you have not yet run BEAUti outside of Eclipse, you may not yet have a $HOME/.beast2 directory. Also, you will need to create the F84 subdirectory inside the $HOME/.beast2/ directory.
This should create the following directory structure: