Interpret omega values for significant log likelihood test?

[Lulu]

Hello!

Sorry for this additional question--I'm very grateful for this discussion group, the manual, and the recent Álvarez-Carretero et al. paper!

I recently performed the branch test on a set of genes. I noticed that some branches I marked as foreground were significant (i.e., significant p-value from log likelihood test) but I noticed the omega value was not greater than 1, indicating positive selection. I was wondering how I should interpret my results--for example if for alt M2 the results were w0=0.0153 and w1=0.6435 while the null M0 w0 for all branches=0.0234, could you infer that the foreground is under relaxed purifying selection (but not strong selection, since it's less than 1?) Because it was a significant log likelihood result, it appears there is evidence that the alt model better explains the data than the null model/reject a single rate of evolution for all branches.

Thank you for your help with this!

[Sandra AC]

Hi there, 

If the result of the LRT led to your null hypothesis being rejected (e.g., "ω does not vary across the branches of the tree"), then the model that best fits your dataset is the one you specified as your alternative hypothesis. If such model was a branch model and the estimated ω ratio for the foreground branch/es was <1, then the gene you have analysed could be under purifying selection for the labelled lineage/s (i.e., nonsynonymous mutations may be fixed at reduced rates than synonymous mutations). Perhaps other people in this group have other comments to add :)

Hope this helps!

S.