Markov chain Monte Carlo (MCMC)

[Bảo Dương Tuấn]

Dear all,

I'm new to BEAST.  Would someone help me to clarify a question from my reviewer?.

Line ---- “ Markov chain Monte Carlo (MCMC) ”:  Was the time signal of the sequence estimated before calculation? What are the principles of model selection?

currently, I am estimating the time of most recent common ancestry (tMRCA) by using a time-scaled phylogenetic analysis of the surface influenza viral via the Bayesian Markov chain Monte Carlo method available in BEAST v1.10.4.

by following youtube and other papers, I used the SRD06 nt substitution model and an uncorrelated relaxed clock with a log-normal distribution to evaluate this data. The bayesian skyline coalescent tree was then selected for this analysis and the Markov Chain Monte Carlo (MCMC) algorithm was applied to produce a relevant post burn-in effective sample size (ESS) of at least 200 for each parameter using Tracer v1.7.1.

The results of ESS is ok then I calculated tMRCA following https://taming-the-beast.org/tutorials/Prior-selection/).

Thank you so much.

 

Best regard,

Duong.

[Daria Budzyńska]

Dear Duong,

As I remember in case of time-scales phylogenies it is necessary to check the presence of temporal signal in the data to avoid any bias. Only after that, if there is a temporal signal in your data, the time-scaled phylogenetic makes sense.You can analyse it with TempEst (https://beast.community/tempest), a really user-friendly tool. I highly recommend you reading Bayesian Evolutionary Analysis with BEAST (https://www.cambridge.org/core/books/bayesian-evolutionary-analysis-with-beast/81F5894F05E87F13C688ADB00178EE00) to become familiar with BEAST.

Best regards,

Daria

--
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/f4713a82-ee1c-4bb2-9bf6-14744578086bn%40googlegroups.com.

--

D. Budzyńska

[Omar Idris]

Does anybody have general idea about why Beast gives wrong tree’s although ESS seems quite good trace is caterpillar like so chain sampling is not the issue. I am trying to figure out why. Other runs in RAXML give correct tree like the genome wide tree or nuclear multiple gene tree under some boot taping options and NJ under composite seems to also agree.

If it helps I am trying to see if a small fragment 493 16s RRNA can give correct tree from other larger gene samples if not why. I found the gene to be quite conserved and very few parsimonious sites not more than 20 among group of related species that are controversial may be 5 plus some populations and 3 out groups. The outgroups always show stable tree under different models but the target species show up in different spots showing lack of support. Is there anything that I can tween Beast options? I do not see much options.

I run 10000000 McMc 10 per cent burnin. All posteriors likelihoods and ESS 8000- 9000 all looks good to me.

Thanks

Thanks.

--
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/f4713a82-ee1c-4bb2-9bf6-14744578086bn%40googlegroups.com.

--

ONI