Sorry to bother you, but I have a quick question concerning the “model averaged probabilities that any two branches have the same dN/dS” output generated by GA-Branch. Can these probabilities be taken as p-values where anything <5% would be considered a significant difference and if so, should a Bonferroni correction for multiple comparisons be used prior to determining which lineages are significantly different?

Thanks!
Daryn

Hi Daryn,

The values reported by GABranch are NOT p-values; the closest analogy are posterior probabilities (or something like 1-p), but even then GA only approximates those under some assumptions. If the value is low for a pair of branches, you can say that the models with high c-AIC scores (credible set) assign different rates to these branches. However, I would just leave it at that -- do not try to interpret those as p-values (because those imply a hypothesis test which is not performed here), and definitely do not use the Bonferroni correction.

Best,
Sergei

Thanks!