Hi Sergei,

I am interested in estimating the errors associated with the dN and dS values that are reported by post_sns.bf.  I think I know how to do it, but I have some questions.

Because the dS and synRate trees are exactly proportional, it seems dS for any branch can be calculated by multiplying the branch's synRate by a constant.

dS = C x synRate


Is this the case?  Should these values really be so simply correlated (I am using the MG94W9 codon model with local parameters).

If the above formula is correct, then I think this means that the error associated with dS can be calculated by multiplying the error for synRate by the same constant.  And the error for synRate can be obtained by taking the square root of the variance in synRate (which I am estimating by nonparametric boostrap).  Thus:

error_dS = C x sqrt(variance_synRate)


Does this sound about right to you?

Thanks!
Jeff

Dear Sergei,

Yes, that makes sense.  The C value should change from replicate to replicate and this variance needs to be accounted for.  I'll have a look at the bootstrapping code and see what I can do.

Thanks!
Jeff

Dear Jeff,

The easiest thing to do is output the numbers of synonymous/non-synonymous/total sites in each replicate as additional columns in the .csv file. I actually need this for another project, so I'll modify the file today and post a notice here...

[Part 2]. Just put up a new build, where the bootstrap analysis will also output the number of synonymous and ns sites for each bootstrap iterate into the .csv file it writes. Hence, to get dS and dN estimates (per branch or otherwise), you can rescale E[syn subs/site] by sites/syn sites and E[nonsyn subs/site] by sites/ns. sites

Cheers,
Sergei