Dear Datamonkeys
We have runned GARD to search for breakpoints in a set of 25 sequences (from a multigene family). But after re-running the program 10 times, we find back 1 or 2 breakpoints but at different sites. They do all fall in the same region, but how can we interpretate the data if it is not consistent? We want to use it for a PAML analysis to determine positive selection with the segmented regions, so how is the best way to proceede?. Could it be a problem with the alignment?
thanks in advance

Hi Maria,

GARD is a randomized algorithm -- it does always give the same answer. Are the breakpoints from the different runs fall within each others' confidence intervals? For instance, if run1 calls a breakpoint at 200, and the confidence interval is 150-250, and run2 calls a breakpoint at 180, then the results are not contradictor, but rather indicate lack of information content in the alignment.

PARRIS (implemented in datamonkey) is our implementation of partitioned selection analysis, but you should be able to do an analogous alignment (albeit with more programming and constant synonymous rates) in PAML.

Sergei