[CodeML- Branch model] 0.0001 and 999.00 dN/dS values at some nodes of the tree.

[Peter Atanackov]

I have been having this problem repeatedly. In this instance I have a set of sequences of a very conservative gene which has a role in maintaining genomic stability distributed across mammals. 

 

I ran the branch model and got the following tree out of it: 

 

It looks okay, except for the ridiculously high or low dN/dS values at some nodes. This is mainly due to the fact, that at these particular nodes the dN value was 0 or sometimes both were 0 (except of course for where the value 999 is, there the dS value was 0).

 

I have no idea why this is happening, is it because no synonymous or nonsynonymous substitutions were detected at these nodes, or is it just something about the algorithm that does this, or are my parameters set wrong? I'm clueless, if anyone has the slightest idea, please enlighten my efforts. 

 

[Ziheng]

if you are using the free-ratios model with a different w for each branch, then yes, lack of synonymous or nonsynonymous substitutions along the branch would cause extreme estimates.  This model is very parameter-rich and its use is discouraged.   You can try model = 2 if you have some idea about what kind of positive selection you are testing and where it is likely to occur.
ziheng