About estimated clock rate in BEAST

[Shota Murakami]

Hi everyone,

I have questions about clock rates estimated in BEAST.

I'm investigating the phylogeny of voles and estimating its divergence times.

I sequenced a part of control regions of mitochondrial DNA.

In my BEAST analysis, I set a calibration point at 2.5 MYA (Lebedev et al. 2007) and estimated clock rates, based on the previous reports: 1.7-1.85%/MYA (Matson and Baker 2001) and 8.5%/MYA (Fedorov and Stenseth 2001). For estimating clock rates, I assumed both the lower limit at 1.7 and the upper limit at 8.5 for prior setting, and estimated parameters with normal distribution assumed.

My first questions are

1) If you don't know proper clock rates for your samples but you have some of different information of clock rates (e.g. 1.7%/MYA and 8.5%/MYA), how do you use the information and estimate clock rates?

2) Is the estimating method performed in my analysis (used different mutation rates as prior information and assumed normal distribution for estimating clock rates) formal?

3) Although many BEAST users estimate clock rates based on previous reports of divergence rates or mutation rates, they don't tend to describe newly estimated clock rates on their papers, why?

I apologize for my poor English skills.

Best,

Shota

[Remco Bouckaert]

Hi Shota,

> 1) If you don't know proper clock rates for your samples but you have some of different information of clock rates (e.g. 1.7%/MYA and 8.5%/MYA), how do you use the information and estimate clock rates?

You can use a parametric distribution (e.g. log normal) and set parameters so that the 95%HPD interval matches that of the clock rate. BEAUti tells you what the intervals are, so you can try a few parameter values till you find a good match.

> 2) Is the estimating method performed in my analysis (used different mutation rates as prior information and assumed normal distribution for estimating clock rates) formal?

Since clock rates must be positive, a log normal distribution is appropriate, and it will resemble a normal distribution inside the 95% HPD when using a small standard deviation. Make sure to set “mean in real space” to true, and start with a mean value of halfway the interval before tuning the standard deviation.

Since you already have a calibration for a node in the tree that possibly is not compatible with the rate from the literature, you want to compare prior and posterior clock rates to see whether this is the case.

Cheers,

Remco

[Shota Murakami]

Thanks, Remco

I'm grateful to your reply and kind helps!

Best regards,

Shota

2018年1月10日水曜日 3時24分02秒 UTC+9 Remco Bouckaert:

[sshweta]

Hi Remco,

Could you please help me out in estimating Divergence time of my plant.

I am using the BEAST2 software.

I have substitution rate 7.1e-3 per year.

Where can I look for divergence time???

[Remco Bouckaert]

> On 2/02/2018, at 9:39 PM, sshweta <shuklas...@gmail.com> wrote:
>
> I have substitution rate 7.1e-3 per year.

Do you know the uncertainty in that estimate? If so, you can estimate the clock rate and put a prior on the clock rate, say a log-normal with a mean of 7.1e-3 and set the variance to match the uncertainty. It is easiest to do this in BEAUti, since it shows the distribution in a graph with some of its statistics.

> Where can I look for divergence time???

The tree height is reported in the trace log and can be seen in Tracer.

If you want to know the divergence time of a subset of taxa, you need to specify an MRCAPrior (click the “+ Add Priors” button in the priors panel in BEAUti) and specify that set of taxa, and you will be able to see the distribution of the divergence time in Tracer.

Otherwise, you can find divergence times in the tree log, which you can load in DensiTree or summarise with TreeAnnotator and load in FigTree.

Remco

[sshweta]

Thank you so much for well explaining.Could you please help me out with one more confusion 

I  have Three TMRCA (Time to Most Recent Common Ancestor) reference points were used to calibrate the clock and estimate the divergence time in rooted time-measured molecular phylogenies of the concatenated data set.

Could anyone help me out how can I use these ages while setting parameter in BEAUTi???

I shall be very thankful!