Dear Sergei,

Before I start, I would like to let you know that I am an absolute beginner. So please don't be annoyed if my questions seem too trivial to you.
I have an alignment of bacterial sequences and I believe there could be some dual coding region in somewhere close to middle of the alignment. I started to look if there are some models in HYPHY which would allow me to check this.  And I found this post where you have a pretty long conservation
( conversation number- num=1149266253/4#4, sorry but I was not allowed to post the URL in this message)

However, as you suggested, do we really need to complicate things with such models ?
I think a relatively simple thing to do is to just look at the rates of synonymous substitutions in one of the frames. If there is indeed dual coding, I should see a much reduced rate of synonymous substitutions in this region. Because synonymous positions in one of the frames would be non-synonymous in the other reading frame.

So I believe I should use dNdSBivariateRateAnalysis.bf to see if there is variation in Ks.
Do you have something to suggest ? I have some technical questions as well but I would like to be sure first that I am doing things the right way.

Thanks and regards

Sankalp

Hi Sankalp,

Sure, the heuristic you suggest (combined with sufficient length with no stop codons in either frame) should work. As an exploration tool, I would suggest that you use FEL on Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login (see Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login) and examine the site-by-site plot of dS for 'islands' of low rates. BivariateCodonRateAnalysis (also implemented on Datamonkey as Evolutionary Fingerprinting) can infer the distribution of site-to-site rates for you, but it won't necessarily be the best tool for looking at contiguous regions of lower dS rates. For that, you may be better off crafting a model with the spatial hidden Markov component (a-la the old Churchill and Felsenstein paper in MBE, 1996 if I recall), that explicitly searches for autocorrelation in dS rates.

Sergei