Infinite log probability when MRCA prior for age is offset

49 views
Skip to first unread message

Guillaume L

unread,
Dec 6, 2021, 12:10:00 PM12/6/21
to beast-users
Hi all,

I fail to run Beast (v2.6.6) for molecular dating with 20 sequences of Primates. In short, the prior is -Inf at initialization, and only when I use an "offset" parameter for the prior distribution of the root age.

The root age (Primates) should be at least 60 My, so I use offset=60, and a Gamma distribution with alpha=2 and beta=5, so the 5% quantile is 61.2 and the 95% quantile is 83.7.

I simplified the analysis to :
- HKY site model;
- strict clock, initial rate of 0.05;
- initial tree computed with UPGMA. But a fixed tree would be fine as well.

My alignment is a simulation with rate 0.05 substitution/nucleotide/My, on a tree with Primates at 70 My.

I get the following error, also when sampling from the prior:

===============================================================================
Start likelihood: -Infinity after 15 initialisation attempts
P(posterior) = -Infinity (was -Infinity)
    P(prior) = -Infinity (was -Infinity)
        P(YuleModel.t:sim000_fsa) = -43.79731979956697 (was -43.79731979956697)
        P(YuleBirthRatePrior.t:sim000_fsa) = -4.605169185987592 (was -4.605169185987592)
        P(ClockPrior.c:sim000_fsa) = -0.8775968828829954 (was -0.8775968828829954)
        P(KappaPrior.s:sim000_fsa) = -1.8653600339742873 (was -1.8653600339742873)
        P(Primates.prior) = -Infinity (was -Infinity)
    P(likelihood) = NaN (was NaN)  **
        P(treeLikelihood.sim000_fsa) = NaN (was NaN)  **
Fatal exception: Could not find a proper state to initialise. Perhaps try another seed.
See http://www.beast2.org/2018/07/04/fatal-errors.html for other possible solutions.
java.lang.RuntimeException: Could not find a proper state to initialise. Perhaps try another seed.
See http://www.beast2.org/2018/07/04/fatal-errors.html for other possible solutions.
    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 java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
    at java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.base/java.lang.reflect.Method.invoke(Method.java:566)
    at beast.app.beastapp.BeastLauncher.run(Unknown Source)
    at beast.app.beastapp.BeastLauncher.main(Unknown Source)
Fatal exception: Could not find a proper state to initialise. Perhaps try another seed.
See http://www.beast2.org/2018/07/04/fatal-errors.html for other possible solutions.
java.lang.RuntimeException: An error was encounted. Terminating BEAST
    at beast.app.util.ErrorLogHandler.publish(Unknown Source)
    at java.logging/java.util.logging.Logger.log(Logger.java:979)
    at java.logging/java.util.logging.Logger.doLog(Logger.java:1006)
    at java.logging/java.util.logging.Logger.log(Logger.java:1029)
    at java.logging/java.util.logging.Logger.severe(Logger.java:1776)
    at beast.app.beastapp.BeastMain.<init>(Unknown Source)
    at beast.app.beastapp.BeastMain.main(Unknown Source)
    at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
    at java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.base/java.lang.reflect.Method.invoke(Method.java:566)
    at beast.app.beastapp.BeastLauncher.run(Unknown Source)
    at beast.app.beastapp.BeastLauncher.main(Unknown Source)

I don't understand what is wrong in my initialisation/prior.
A clock rate of 0.05 and a Primate age of 60 My seems realistic to me... Or should I also tune the rate of the site model?

Thanks a lot for your help.

Guillaume
Reply all
Reply to author
Forward
0 new messages