zhangchicool / speciesnetwork Goto Github PK

TODO

As @zhangchicool mentioned to me, SpeciesNetwork has problems with many different gene trees. One simple way around this might be to use a tree selector class, with inputs List<TreeInterface> trees and IntegerParameter index (1-dimensional) and behaving as if it was trees.get().get(index.getValue()) for TreeInterface purposes. Having one of these for each gene tree, and uniform integer operators to change their values of index, trivially implements a reversible jump method that groups genes into groups with identical gene trees. (This may have bad convergence, but I'd be happy to at least try it out.)

However, this tree selector class would not be a useful subclass of Tree, so it would not work as input for GeneTreeInSpeciesNetwork via EmbeddedTree. Given the possible issue #4 with embedding initialization and this use case, might it be prudent to drop EmbeddedTree and permit GeneTreeInSpeciesNetwork to accept a TreeInterface as its geneTree input and take over the responsibility to track the Embedding in its own methods?

The redundant blocks added look like this:

<init id="RandomTree.t:47" spec="beast.evolution.tree.RandomTree" estimate="false" initial="@Tree.t:47" taxa="@47">
    <populationModel id="ConstantPopulation0.t:47" spec="ConstantPopulation">
        <parameter id="randomPopSize.t:47" name="popSize">1.0</parameter>
    </populationModel>
</init>

After wondering why I get so many NullPointerExceptions,

	at speciesnetwork.GeneTreeInSpeciesNetwork.recurseCoalescentEvents(GeneTreeInSpeciesNetwork.java:158)
	at speciesnetwork.GeneTreeInSpeciesNetwork.recurseCoalescentEvents(GeneTreeInSpeciesNetwork.java:158)
	at speciesnetwork.GeneTreeInSpeciesNetwork.recurseCoalescentEvents(GeneTreeInSpeciesNetwork.java:158)
	at speciesnetwork.GeneTreeInSpeciesNetwork.recurseCoalescentEvents(GeneTreeInSpeciesNetwork.java:158)
	at speciesnetwork.GeneTreeInSpeciesNetwork.recurseCoalescentEvents(GeneTreeInSpeciesNetwork.java:158)
	at speciesnetwork.GeneTreeInSpeciesNetwork.recurseCoalescentEvents(GeneTreeInSpeciesNetwork.java:158)
	at speciesnetwork.GeneTreeInSpeciesNetwork.update(GeneTreeInSpeciesNetwork.java:117)
	at speciesnetwork.GeneTreeInSpeciesNetwork.computeCoalescentTimes(GeneTreeInSpeciesNetwork.java:96)
	at speciesnetwork.MultispeciesCoalescent.calculateLogP(MultispeciesCoalescent.java:90)
	at beast.core.util.CompoundDistribution.calculateLogP(CompoundDistribution.java:114)
	at beast.core.MCMC.propagateState(MCMC.java:496)
	at beast.core.MCMC.doLoop(MCMC.java:383)
	at beast.core.MCMC.run(MCMC.java:353)
	at beast.app.BeastMCMC.run(BeastMCMC.java:357)
	at beast.app.beastapp.BeastMain.<init>(BeastMain.java:113)
	at beast.app.beastapp.BeastMain.main(BeastMain.java:659)

I noticed the assertions in

I don't know how traversal numbers should behave and what the getDirection function is supposed to do,

        final int traversalNodeNumber = speciesNetworkNode.getTraversalNumber();
        final Integer nextSpeciesBranchNumber = geneTree.embedding.getDirection(geneTreeNodeNumber, traversalNodeNumber);

but the calls above that look reasonable.
@zhangchicool I sent you my XML files that produce this issue, in case that might help track down the problem because it's a user error.

I'm trying to set some constraints in gene trees, but in order to get a non -Inf likelihood I have to also set compatible starting trees, as we would do in e.g. BEAST or StarBEAST. But I'm struggling to accomplish that. Basically what I did was follow a (working) xml file generated for BEAST and added a couple of <init> tags with spec="beast.util.TreeParser" inside <run>, just before the <distribution id="posterior"> tag. Everything seems to be ok, except that I'm getting this error:

Error 110 parsing the xml input file

validate and intialize error: class beast.util.TreeParser cannot be cast to class speciesnetwork.EmbeddedTree (beast.util.TreeParser and speciesnetwork.EmbeddedTree are in unnamed module of loader beast.util.BEASTClassLoader @7229724f)

Am I in the right path? Is there another class I should use for that? And how?

Thank you so much!

It should put the mean, median, and HPD intervals of node heights and gamma probs (if apply) in the attributes.

Dear developers,

I am wondering whether invariant sites affect the results.
Although I read some suggestions that I should include invariant sites in the google group of "beast-users", I want to know whether that case is the same to SpeciesNetwork.

I randomly collected 100 successive SNPs in each chromosomes from VCF file.
Does it affect the final result in SpeciesNetwork?

Best regards.

As far as I can see, the code that sets ν and λ from the input parameters