artpoon/bellejs

Generate BEAST XML analysis files using a BEAUTI-like JavaScript interface

bellejs's Issues

UCLN has extra XML elements

	<rateStatistic id="meanRate" name="meanRate" mode="mean" internal="true" external="true">
		<treeModel idref="treeModel"/>
		<discretizedBranchRates idref="branchRates"/>
	</rateStatistic>
	
	<rateStatistic id="coefficientOfVariation" name="coefficientOfVariation" mode="coefficientOfVariation" internal="true" external="true">
		<treeModel idref="treeModel"/>
		<discretizedBranchRates idref="branchRates"/>
	</rateStatistic>
	
	<rateCovarianceStatistic id="covariance" name="covariance">
		<treeModel idref="treeModel"/>
		<discretizedBranchRates idref="branchRates"/>
	</rateCovarianceStatistic>

Running BEAST v1.10.4 on the generated XML file

Getting the following error when running BEAST on the generated YFV.xml file:

SEVERE: Parsing error - poorly formed BEAST file, YFV.xml:
Object with Id=2009_TVP11767_Trinidad already exists

Error thrown at: dr.xml.XMLParser.convert(Unknown Source)

Parse tip dates

In BEAUTi, clicking on the "Use tip dates" checkbox activates the other inputs, such as the "Parse Dates" button.
Clicking on "Parse dates" creates a modal window where the user specifies how the app should extract date information from the sequence headers, e.g., delimiter character, position of date in tokens.

Generate XML from current state of interface

When user clicks "Generate BEAST file" button at the lower right of the page, it should trigger an event handler that polls the various inputs and data objects (e.g., alignment) and use that information to generate an XML file. Might need to do a JSON to XML conversion.

Interface throws error when attempting to make XML a second time

Failed to locate branchRates element in rateStatistic

Starting Tree tab?

Should we include this tab?

To change tree type to cluster Tree, and cluster tree type to neighbourjoining. (working off my previous year's BEAST lab)

First attempt at skyline run

Creating the tree model, 'treeModel'
  taxon count = 71
  tree height = 152.9034811119163
Bayesian skyline plot: 50 linear control points
Fatal exception: Dimension of population parameter must be one greater than dimension of group size parameter.
java.lang.IllegalArgumentException: Dimension of population parameter must be one greater than dimension of group size parameter.
	at dr.evomodel.coalescent.BayesianSkylineLikelihood.<init>(Unknown Source)
	at dr.evomodelxml.coalescent.BayesianSkylineLikelihoodParser.parseXMLObject(Unknown Source)
	at dr.xml.AbstractXMLObjectParser.parseXMLObject(Unknown Source)
	at dr.xml.XMLParser.convert(Unknown Source)
	at dr.xml.XMLParser.convert(Unknown Source)
	at dr.xml.XMLParser.parse(Unknown Source)
	at dr.app.beast.BeastMain.<init>(Unknown Source)
	at dr.app.beast.BeastMain.main(Unknown Source)

ucld.mean has a unique prior

If we select the uncorrelated lognormal model and do not parse tip dates, ucld.mean has no prior distribution (appears as Fixed value, value=1).

However, this setting appears if you set tip dates:

<ctmcScalePrior>
	<ctmcScale>
		<parameter idref="ucld.mean"/>
	</ctmcScale>
	<treeModel idref="treeModel"/>
</ctmcScalePrior>

Debugging XML (JC, strict clock, constant coalescent, dated tips)

Mar 30, 2021 9:33:54 PM dr.app.beast.BeastMain <init>
SEVERE: Parsing error - poorly formed BEAST file, belle-jc-strict-dated.xml:
Error parsing '<scaleOperator>' element with id, 'null':
Scale operator can only be used on parameters constrained to be strictly positive or negative (clock.rate)

Error thrown at: dr.inferencexml.operators.ScaleOperatorParser.parseXMLObject(Unknown Source)

Need to bind contents of Priors tab to other tabs

@ewong347 suggested creating persistent Prior objects for every possible model combination, which is a good idea (we can also have a many-to-one mapping when models share the same prior).

Comparing BEAUti and Belle XML outputs

JC, strict, constant coalescent
Belle has extra entry:

        <!-- Statistic for time of most recent common ancestor of tree               -->
	<tmrcaStatistic id="age(root)" absolute="true">
		<treeModel idref="treeModel"/>
	</tmrcaStatistic>

Minor differences in HKYModel

BEAUti:

	<!-- The JC substitution model (Jukes & Cantor, 1969)                        -->
	<HKYModel id="jc">
		<frequencies>
			<frequencyModel dataType="nucleotide">
				<frequencies>
					<parameter id="frequencies" value="0.25 0.25 0.25 0.25"/>
				</frequencies>
			</frequencyModel>
		</frequencies>
		<kappa>
			<parameter value="1.0"/>
		</kappa>
	</HKYModel>

Belle:

	<!-- The HKY substitution model (Hasegawa, Kishino & Yano, 1985)             -->
	<HKYModel id="hky">
		<frequencies xmlns="http://www.w3.org/1999/xhtml">
			<frequencymodel dataType="nucleotide">
				<frequencies>
				  <parameter id="frequencies" value="0.25 0.25 0.25 0.25"></parameter>
		  	</frequencies>
	  	</frequencymodel>
  	</frequencies>
		<kappa xmlns="http://www.w3.org/1999/xhtml">
			<parameter id="kappa" value="1.0" lower="0.0"></parameter>
		</kappa>
  </HKYModel>

Empirical base frequencies

If we estimate base frequencies:

                <frequencies>
                        <frequencyModel dataType="nucleotide">
                                <frequencies>
                                        <parameter id="frequencies" value="0.25 0.25 0.25 0.25"/>
                                </frequencies>
                        </frequencyModel>
                </frequencies>

If we set frequencies to empirical:

                <frequencies>
                        <frequencyModel dataType="nucleotide">
                                <alignment idref="alignment"/>
                                <frequencies>
                                        <parameter id="frequencies" dimension="4"/>
                                </frequencies>
                        </frequencyModel>
                </frequencies>

Test XML: HKY, estimated frequencies, UCLN clock, constant coalescent

Using TreeDataLikelihood
  Branch rate model used: discretizedBranchRates
Mar 30, 2021 10:11:15 PM dr.app.beast.BeastMain <init>
SEVERE: Parsing error - poorly formed BEAST file, belle-hky-ucln-dated.xml:
Object with idref=ucld.stdev must not have other content or attributes (or perhaps it was not intended to be a reference?).

Error thrown at: dr.xml.XMLParser.convert(Unknown Source)

Enable user to modify prior settings (hyperparameters)

Clicking on row of priors table should bring up modal window where user can view and modify hyperparameters, e.g., mean and standard deviation of normal distribution.
Let's not worry about visualizing the distribution for now.

XML fails to run

Attempting YFV FASTA file with tip dates and UCLN clock.

Read alignment: alignment
  Sequences = 71
      Sites = 654
   Datatype = nucleotide
Site patterns 'patterns' created from positions 1-654 of alignment 'alignment'
  unique pattern count = 280
Mar 29, 2021 11:30:52 PM dr.app.beast.BeastMain <init>
SEVERE: Parsing error - poorly formed BEAST file, test.xml:
Object with idref=taxa must not have other content or attributes (or perhaps it was not intended to be a reference?).

Error thrown at: dr.xml.XMLParser.convert(Unknown Source)