Questions about the number of states, burnin, Treeannotator, Tracer and Logcombiner

[Jackson Lima]

Hello BEAST Users,

I have some questions about the process for burnin trees and log.

I need to conduct a BSP analysis of some genes of a virus, to get the ESS recommended (> 200) were required 3 different runs, each run had 30,000,000 states and with sampling every 100, which theoretically would generate a LOG file TREES and with 300.000 states, ok?
By using LOGCOMBINER to get a file with a mix of 3 runss, with a burnin of 10% (30,000) both to LOG file as the file to TREES, so the output files will LOGCOMBINER 810.000 states, correct?
But that's not what happens, the LOG file of each race, when analyzed separately in TRACER, show me 30,000,000 states and a burnin of 3,000,000 and a combined analysis of three races with 81,000,300 states, there should be 81,000,000? Why is there such 300 more? If I join the same 3 files in LOGCOMBINER, opening the TRACER and analyze, it shows 90,000,200 states and a burnin of 9,000,020. By now it adds 200 states? In TRACER he added 300.

I'm lost in the sampling process that I set in BEAUTI because I think I configured one way, but actually happens otherwise.
It saves the 30,000,000 states that the race has or every 100 as configured?

To TREES file is the same situation?

It saves the state 30,000,000 or makes sampling every 100?

Considering the TREES file, he saved 30,000,000, I realized a merge of 3 runs in LOGCOMBINER with a burnin of 3,000,000, which would give me a file with 81,000,000 states, correct?
To build a tree, went to TREEANNOTATOR, configured as the tutorials available on the website (Mean Heights, Maximum Clade credibility tree, posterior probability limit 0.0) and burnin of 8,100,000 and put to run, after he read the trees, the following message: "Finished - Quit program to exit.
No trees to use: burnin too high "

And now, the number of states is not the one that set up? What should I do?

If you use a number, which I think is what setup (30,000,000 states and sampled every 100, giving a file with 300.000 states), joining in the 3 LOGCOMBINER, give a file with 810.000 states and build the tree in TREEANNOTATOR with a burnin of 81.000, the program can not, because the last part of the program before finishing all the construction reaches 75% it crashes, it consumed all the ram (the pc has 32 GB of RAM and TREEANNOTATOR is configured to use 31 GB ).
Someone could explain why this occurs? Any tips or help to solve my problems?

I also noticed that when running TREEANNOTATOR by command line, the following message appears: "JRI not available. Bivariate Using Java attributes," This could explain the excessive consumption of ram?

Well those were the questions I have, if anyone can help me I appreciate it.

Best Regards,

Jackson Lima

PS: Everything was done with the BEASTv1.7.4.