BEAST vs. r8s

[Julie Dragon]

I am new to this type of analysis and have some basic questions. I
would like to use the existence of two Hawaiian Island endemics to
date the divergence events on my phylogenetic tree. I have run
analyses in both r8s and BEAST and get very different node ages across
the tree. In r8s, with both Langely-Finch and penalized-likelihood
algorithms, using constrained or fixed ages for the nodes used to date
the tree, I get the same dates across the tree (ingroup=3.1 mya).
Using a relaxed lognormal clock in BEAST, with a fixed node age
(normal distribution) I get much older dates (ingroup=13 mya). Using
a constrained node age (uniform distribution) I get much younger dates
(ingroup= 0.0041 mya). If I use the mean substitution rate identified
in the Langley-Finch analysis for the meanRate (normal distribution)
in BEAST, I get even older node ages. I freely admit that I do not
understand the differences in how these two programs work. Can anyone
explain the fundamental difference in these kinds of analyses and why
I might get such different results?

Any advice is much appreciated!

[Simon Ho]

Hi Julie,

The methods implemented in BEAST and r8s are fundamentally quite
different, but the differences that you have observed seem to be very
large. I suspect that there might be a problem with at least one of
your analyses. It is difficult to tell what might be going wrong, but
you could start by considering the following:
1) Is the tree topology estimated by BEAST the same as the fixed
topology being used for r8s?
2) Are you using a strict or relaxed clock in BEAST?
3) What sort of calibrations are you using - do they provide
sufficient information to be able to estimate rates/dates?
4) What was the optimised smoothing parameter in your PL analysis? If
it was very high, then that means your data are clocklike (which is
probably the case if LF and PL are giving the same results), and a
strict clock in BEAST should yield similar results.
5) Are you analysing intraspecific or interspecific data, or a mixture
of the two? This will affect your choice of tree prior in the BEAST
analysis.
6) Have you tried a BEAST analysis with a strict clock, using the rate
obtained from the LF analysis in r8s? This should give very similar
date estimates to the r8s analyses.

Cheers,
Simon

[Julissa Roncal]

To All Beast users:

 

I want to use a "user target tree"  in TreeAnnotator to summarize node ages obtained after runing BEAST. This is a strict consensus tree obtained from PAUP.  

 

#NEXUS

Begin trees; 
 tree Strict = [&U] (Sere_repe,((((((((((Lacc_aust,Howe_fors),(Ptyc_maca,Carp_macr)),Loxo_rupi),Dyps_lute,((Dict_albu,Rhop_lede),Onco_tigi)),(Neon_wats,Eute_prec),(Leop_pulc,Pela_henr)),((((G.max_peru,(G.calyp,(G.mac_acau,G.mac_macr),(G.caman,G.trigl),G.multi,G.pinn1262,G.pinn1261)),((G.orb_chir,((G.undat,G.juss1269),G.orbig)),G.tenui,G.poepp,G.cunea,G.stric,G.lanat,(G.brong,(G.str_subm,G.pasce)),G.longe),G.dever,G.inter,(G.eleg1316,(G.scho1307,G.brev1319),G.pauc1315)),(((Cagy_ghie,Cagy_cost),Cano_occi,Cano_riva),Cano_plum),Aste_mart),(((Phol_dact,Phol_pulc),Phol_syna),Welf_regi))),Mani_sacc),(Wett_hirs,Socr_exor)),((Scle_mann,Oran_laut),Podo_bart),((((Alla_aren,Syag_smit),Atta_alle),Becc_mada),(Elae_olei,((Gast_cris,Acro_acul),Astr_para)),Rein_simp,Roys_oler),(((Syne_fibr,Gaus_maya),Cham_micr),Hyop_lage)),(((Cero_quin,Rave_rivu),(Phyt_aequ,Apha_nata)),Pseu_vini)),(Thri_radi,Cary_miti));
 

end;

 

But I get the following error message:

 

Error Parsing Target Tree: Missing ',' in tree in TREES block
Total trees read: 10001
Ignoring first 1000 trees.
Reading user specified target tree, C:\Documents and Settings\Adm\My Documents\Aarhus DK\Geonomateae DNA sequences\comb div time_CONedited.tre
Finished - Quit program to exit.

Does anyone know what kind of format my "user target tree" must have?

 

Thank you very much in advance

 

Julissa Roncal