divergence time estimate using substitution rate in star beast

[Zhang Dz]

Dear all,

I am using some introns to estimate divergence time implement in star beast for 3 species. I have no calibration time, so I chose the strict clock with an value of 0.0033 (0.33 per million years).

Seems like the resulted tree is good, but the divergence time is extremely low. Did I misuse star beast for divergence time estimate? Any comments would help a lot!

Thank you all!

Dezhi

[Zhang Dz]

I forgot to post the tree, here is it. The divergence time is so small, sth is wrong...

在 2018年1月4日星期四 UTC+8上午11:30:36，Zhang Dz写道：

[Huw A. Ogilvie]

Hi Dezhi,

Did you link the clocks for all loci first, or if not did you set each locus individually to 0.0033? As long as all the clocks are fixed (so "estimated" is not ticked) at that rate, a fixed clock should worked for calibration. If you upload the XML or send it to me, I can check and try to work out what's going on.

- Huw

[Zhang Dezhi]

Dear Huw,

Thank you very much for your help. I linked the clocks according to your suggestion. The new tree seems much better, but the root divergence time is still much lower than expected. It is only one quarter of divergence time of mitochondrial sequences. How can I explain this?

Best wishes,

Dezhi

在 2018年1月4日星期四 UTC+8下午12:00:21，Huw A. Ogilvie写道：

[Huw A. Ogilvie]

Divergence times are expected to be more recent than coalescent times due to the waiting time to coalescent. 4x does sound like a large difference, but that would be expected if your species divergence times are fairly recent, and the effective population size (scaled by generation time) was large. Say if the root height was only 1 million years ago, and the effective population size is around 3 million individuals, then the expected coalescence time of the mitochondrial genomes will be 4 million years ago assuming an annual generation time.

On the other hand, if your divergence times are much older than this, or if the effective population size (scaled by generation time) is much smaller, that would suggest some kind of error in the analysis. An obvious source of error would be if the suspected intron substitution rate is too slow or the estimated mitochondrial rate is too fast. If you are more confident in the mitochondrial rate, you could include the mitochondrial sequences in your analysis and fix the rate of those genes, and estimate the intron clock rate rather than fixing them. In the most recent version (StarBEAST2 v14) this would require some thought about the specification of the model. For instance you could:

- link the mitochondrial tree, site and clock models (only necessary with more than one mitochondrial locus)
- link the intron clock models
- set the mitochondrial gene ploidy to 0.5
- untick "estimate" from the mitochondrial *site model* substitution rate
- fix the mitochondrial *clock model* rate at the known value
- tick "estimate" for the intron *clock model* rate, and set the intron clock rate prior to 1/X, or to a log-normal distribution (with a mean in real space around the suspected intron substitution rate)

Regards,

- Huw