divergence time estimation of intra- and inter-species in a single tree

[Jiao]

Dear all,
I am using BEAST for dating of a phylogenetic tree, which contained
both 36 haplotypes of a species (Cordyceps sinensis) and other 20
species (Cordyceps spp) as outgroups. Within C. sinensis, a
significant divergence between different geological populations was
observed in phylogenetic analysis. Thus, I am interested in both the
divergence time within C. sinensis and the time framework of these
Cordyceps spp. I calculated the age of nodes using 'Relaxed Clock:
Uncorrelated Lognormal' model, 'Speciation: Yule Process' as Tree
Prior, and a normal distributed prior for the TMRCA of C. sinensis and
its sister clade. The figures are shown in:
phylogenetic tree: http://photos.i.cn.yahoo.com/sduxiaoren/5b9f/a048.jpg/
molecular clock result: http://photos.i.cn.yahoo.com/sduxiaoren/5b9f/5577.jpg/

I found that the mean divergence time (~1.45 Mya) within C. sinensis
seems to be enlarged and even bigger than the split time between some
speices, although the genetic divergence level within C. sinensis were
less than that between those spp. On the other hand, if I use two
representative haplotypes of C. sinensis combined with 20 outgroups,
the mean divergence time between these two haplotypes were extremely
reduced to 0.45 Mya.

So my questions are:
- From the point of view of molecular clock calculation, is it
acceptable to estimate divergence time of both within a speices and
among spp. in a tree on the same time? (All 36 haplotypes and 20
outgroups)
-- If it is acceptable, should I reduce the number of outgroups or
reduce the haplotype number of C. sinensis to get a more reliable
timeframe coordinated with the branch length of phylogeny?
-- If not, is it acceptable to split all samples to '36 haplotypes +6
outgroups' and '1 haplotype +20 outgroups' to calculate the divergence
time within species and among spp. separately?


Thanks a lot and I will be eagerly expecting your response.

Jiao

[Simon Ho]

Dear Jiao,

In your analysis of 2 haplotypes + 20 outgroups, are you choosing the
2 most divergent C. sinensis haplotypes?

There are some problems with running an analysis of 36 haplotypes + 20
outgroups. The two major problems are:
(1) Neither a Yule prior nor a coalescent prior in BEAST would be
appropriate;
(2) You are likely to observe a different rate within the C. sinensis
clade compared with the interspecific branches.

The best option would be to run an analysis with 1 haplotype + 20
outgroups using a Yule prior.
Regarding the intraspecific analysis, ideally you would only analyse
the 20 haplotypes, using a coalescent prior, calibrated within C.
sinensis. If you have no choice but to include the sister species for
calibration, then perhaps you could try running the analysis with a
coalescent prior placed only on the C. sinensis clade. Let me know if
you'd like to try this and I can send you an example of an input file
that specifies this.

Simon

[Giovanni Zecca]

Dear Simon,

I'm quite new in using BEAST. I have a situation similar to the one
posted by Jiao.

1)I'm analysing data from several stands of Euphorbia spinosa
collected in different geographic locations (mainly from Italy). The
geological history of the area and different approaches (nuclear
sequences, cpSSR and ISSR) suggest a division within Italian E.
spinosa (i.e. the species seems to be split into two groups) and the
presence of a large secondary contact zone, characterized by the
presence of hybrids. The species is not well studied, for this reason
all my calibration points are referred to outgroup species. Within E.
spinosa I'm not interested in estimating other than the main division.

2)On the other hand I'm also interested in dating divergence time
between some couples of outgroup species that live in the same area.

3)Phylogenetic tree is already well resolved at the species level.

So far I ran my analysis excluding hybrid sequences, using Yule
Process as Tree Prior and Relaxed Clock (Uncorrelated Lognormal
model). Results are plausible, concurring with the geological history
of the study area and with the history of other known plants living
in the same area.

Is that enough? I think my situation should be closer to a
phylogenetic study than a population one. Am I steeply wrong?
In your opinion, where are, if any, possible problems? Can my dates
estimation be unduly biased? Is the direction of an eventual bias
predictable (i.e. over or under estimation)? Should I try the solution
you suggested to Jiao (a coalescent prior on the E. spiniosa clade)?
If yes, could you please send me an example of an input file that
specifies this.

Many thanks,

Giovanni.

[alexei....@gmail.com]

Dear Jiao,

The Yule prior assumes that the birth rate of new lineages is the same
everywhere on the tree. In your case it would assume that the birth
rate of new lineages within C. sinensis is the same as between
different species in your outgroup. This assumption in clearly
inappropriate in your case. However you may want to try to isolate
exactly what is causing the problem. The relaxed clock also allows
quite some scope for changes in rate from branch to branch -- so one
possibility is that the relaxed clock is assigning slow rates within
C. sinensis so that longer branches (in time) can be assigned, in
doing so fitting the Yule prior better. To get some insight into
whether this is the case you can:

(1) look at the rates across branches
(2) try the same analysis with a strict clock to see how that changes
you estimate of the ingroup root height.

It is always good to investigate your data thoroughly and see how
different models and priors affect your estimates.

Having said this, Simon Ho's suggestions to try to separate out your
population-level analysis from your phylogenetic analyses is a good
one.

Cheers
Alexei

[hugo...@gmail.com]

Dear Beast users,

Much in line with previous postings in this topic, I'm interested in
evaluating branching times in a clade that is perhaps more consistent
with intraspecific levels of divergence, although I only have
calibration points for species external to that clade. Therefore, I
would be interested in applying a Yule prior to the other species and
a coalescent prior to the clade of interest. I'd be much appreciated
if someone could send me a file on how to do that.

Best,
Hugo

[dstre...@gmail.com]

Dear all,

I'm new to Beast and I think I have a similar situation to those
described in the previous posts regarding combined analysis of inter-
and intra-specific data. I am working on a large (~450 sequences of
about 600bp each) dataset of time-stamped viral sequences which
technically comprise a single viral species, but form ~20 distinct
genetic and antigenic lineages.

I would like to estimate (1) the divergence times for each lineage
(possibly more similar to an inter-specific analysis) and (2)
demographic rates such as effective population size and population
growth rate for each lineage (an intra-specific analysis). Ideally, I
would run this in a single analysis as described above, but I imagine
that this would get quite complicated if different models of
population dynamics (constant growth, exponential growth etc) are
appropriate for different lineages. If I am not especially interested
in dating the really deep nodes or the TMRCA for the entire species,
would it be valid to simply analyze each lineage independently under a
coalescent prior and use the treemodel.rootheight as the date for the
origin of each lineage? Or is there a more elegant way to do this
given that I would have to repeat the analyses 20 times (once for each
lineage)?

If anyone could give me some insight into the most appropriate
strategy for this type of analysis I would greatly appreciate it.

Cheers,
Daniel

On Jan 30, 2:47 pm, "hugo...@gmail.com" <hugo.ga...@gmail.com> wrote:
> Dear Beast users,
>
> Much in line with previous postings in this topic, I'm interested in
> evaluating branching times in a clade that is perhaps more consistent

> withintraspecificlevels of divergence, although I only have