beast terminate with error

[Rebecca Bengtsson]

Hi all,

 

I am running beast to perform divergence dating. I am working with microbial genomic data and have provided core genome SNPs alignment into beauti. To account for ascertainment bias I have also provided invariant sites in the xml file. When I run beast the following error occurs and my run is terminanted.

 

 

 

Start likelihood: -378.1335540171007

Warning: Overwriting file /Users/rjbeng/Documents/Shigella/data/flexneri/beast/subsample/flexneri_beast.log

         Sample      posterior     likelihood          prior

              0      -378.1335            NaN      -378.1335 --

Fatal exception: n must be positive

Warning: Overwriting file /Users/rjbeng/Documents/Shigella/data/flexneri/beast/subsample/flex_beast.trees

java.lang.IllegalArgumentException: n must be positive

            at beast.util.MersenneTwisterFast.nextInt(Unknown Source)

            at beast.util.Randomizer.nextInt(Unknown Source)

            at beast.evolution.operators.UniformOperator.proposal(Unknown Source)

            at beast.core.Operator.proposal(Unknown Source)

            at beast.core.MCMC.propagateState(Unknown Source)

            at beast.core.MCMC.doLoop(Unknown Source)

            at beast.core.MCMC.run(Unknown Source)

            at beast.app.BeastMCMC.run(Unknown Source)

            at beast.app.beastapp.BeastMain.<init>(Unknown Source)

            at beast.app.beastapp.BeastMain.main(Unknown Source)

            at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)

            at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)

            at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)

            at java.lang.reflect.Method.invoke(Method.java:498)

            at beast.app.beastapp.BeastLauncher.run(Unknown Source)

            at beast.app.beastapp.BeastLauncher.main(Unknown Source)

Fatal exception: n must be positive

 

BEAST has terminated with an error. Please select QUIT from the menu.

I have also performed the same analysis with sampling from prior and beast seems to run okay. I am not sure what is happening and would greatly appreciate some help.

 

I have attached the log output file

 

Many thanks

Rebecca

[HS]

Hi Rebecca,

It is clear that likelihood can't be NaN. But to find out why the likelihood is NaN you should provide more info. Maybe it is good to share the priors tab of beauty or the priors part of xml file.

Best,

Hovhannes

[Danushka Weerasekera]

still its appearing an error massage

--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beast-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/beast-users/7df7c46b-7a8e-4db0-9322-01fed985eec3%40googlegroups.com.

[Remco Bouckaert]

Hi Lucy,

The error message suggests the a UniformOperator has a parameter with zero dimension, which should never happen and suggests perhaps something went wrong when setting up constant sites. 

The log file you posted shows the model, but has not likelihood in it which suggests the log was for a sample from the prior (and the likelihood will be NaN when sampling from the prior). 

If you can make the XML available, I can investigate further.

Cheers,

Remco

[Remco Bouckaert]

Hi Danushka,

Can you give more details on what seems to be the problem? I can import the alignment you posted and run it with a HKY model, strict clock and Yule prior (as in the attached) without any issues.

Remco

[Rebecca Bengtsson]

Hi Hovhannes, 

I have managed to solve the issue by simply restarting beauti, entering all the information again and saving to a new xml file. I think the issue may have been caused by the fact that whenever I reopen a xml file in beauti the alignment file appears to be missing from the partitions window. I have no idea why this is happening, but restarting everything (inputing exactly the same priors) from scratch appears to solve the problem.

Many thanks

Rebecca