Closed Edison2021 closed 4 years ago
Dear @Edison2021,
What this means (I think:) is:
Assume you have a tree with many branches and you divide them into separate groups: G1, G2 and G3. Then you could run aBSREL
on G1 first, find, for example, that there is no selection on that set, and then proceed to test G2 in separate run, and so on.
This approach might not be statistically valid, specifically, it is more likely to find false positive results that would be indicated by the p-value. That's because the multiple hypothesis correction would be applied to each group independently, whereas the valid approach would apply the correction to all groups together, making the test more conservative.
So instead you should decide what you are going to test a priori and run the test once
Best, Sergei
Dear Edison,
Because the number of possible branch sets within a phylogenetic tree is combinatorially large, and performing an analysis with a different branch set constitutes a new statistical test, you may need to take into consideration multiple testing correction if the number of branch sets you're analyzing is high.
Best, Steven
Hi Sergei and Steven Thanks for your information! If I have a tree with four separate groups: G1-G4. May i run absrel separately as follows Assign the G1 to foreground, G2, G3 and G4 as background Assign the G2 to foreground, G1,G3 and G4 as background Assign the G3 to foreground, G1, G2 and G4 as background
Thanks Edison
Dear @Edison2021,
There is no concept of foreground or backbiting in aBSREL. You may choose to test only a subset of branches, and ignore the rest. Can you elaborate on what evolutionary hypotheses you are testing?
Best, Sergei
Hi Sergei I have 40 species that can be separated into 4 groups and each group has a specific trait, I suppose some genes are under selection at ancestral node (the node leading to species sharing trait) contributing to origin of the trait. So i would like to know what genes are.
Best Edison
Dear @Edison2021,
If I understand you correctly, you want to test for selection along the purple branches of a tree that looks like this
There are a few things you could do.
Run aBSREL
only on the 4-branch purple set. This will allow you to test whether or not the ancestral branches to your four groups experienced positive diversifying selection.
Run FADE
on each of the four groups separately. This will test for evidence of directional selection in specific clades
Run RELAX
in group mode (see https://mbio.asm.org/content/10/6/e02524-19/article-info for an example) to determine whether or not selective pressures differ between the four clades. The four groups would be each of the colored branch sets.
Best, Sergei
Hi Sergei Thanks very much for the detailed methods that will be very helpful for my project.
Best Edison
Dear @Edison2021,
No worries. Let me know if need further guidance. How many different genes are you testing?
Best, Sergei
~2500 genes. Not so many single-copy genes were found.
Edison
Hi The 'Rules of thumb for aBSREL use' in the tutorial mentions "we strongly discourage performing multiple separate tests for selection on different branch sets." I am a little confused with the comment, Could you please explain more?
Best Edison