Ignoring monomorphic site (-0) or not?

[PeiWeiSun]

Hello, I have a problem of using -0 or not

I know that by adding -0, mutation rate will be ignored and parameters need to be calibrated by other known parameter

however, the ranking of the likelihoods to my tested models are different if I add -0 or not

My question is, is there any suggestion to ignore or not to ignore monomorphic sites when optimizing observed SFSs?

Thanks for any helps!

[Laurent Excoffier]

Hi,

the discrepancy suggests that you may have problems fitting the total number of polymorphic sites with one or several models. Also the parameter that you fix when using -0 options should be quite informative, like an ancient divergence time or a known ancestral population size.

Also take care of correctly computing the number of monomorphic sites. If this is well done and the mutation rate is well estimated, there is no reason to use the -0 option.

Finally, without the -0 option, try to use the -l and -L option with a reference parameter (look in the manual on how and why to do that)

best

laurent

[PeiWeiSun]

Hello Dr. Laurent

Thank you for your response and suggestions.

I read the manuscript and learned that after the specific ECM cycles set using -l option, only polymorphic sites will be estimated based on a reference parameter. If I and right, without using all sites (monomorphic and polymorphic), the benefit of using monomorphic sites solely is decreasing computational requirements. Or Are there any other benefits of using -l option? 

Thank you for your help

Pei-Wei

laurent....@gmail.com 在 2023年6月23日 星期五晚上7:42:29 [UTC+8] 的信中寫道：

[Shailesh Kumar Dhudabhai Desai]

Hii, 

sorry for replying to exisitng thread, but my question is related to this. 

I have vcf files with biallelic sites only (SNPs), and then using the eastSFS, i converted it into SFS data. 

i am trying to estimate the divergence between two species, and other parameter, but i have mutation rate from other study. 

I understand that if we use -0, we need some external reference parameter like divergence or Ne size, as with -0 we can not use mutation rate, 

but just want to be sure that can i trust my monophoric sites, for this kind of data, as i want to estimate split time. 

I look forward for any help. 

Thank you.