Hello all,
I've detected selection using different methods. I am here showing example CLR method.
What I did is I selected position that have 1% highest CLR value (see picture). and crossed the position with the annotation and I got 1200 genes.
imagine that MYB transcription factor was in the list of candidate region that are under selection.
There is a probably several hundreds of MYB in the genome, so it's very likely that I can find them in the peaks by chance.
How would you really say this gene was under selection if it's at the end is a matter of chance?
Ideally you would need to simulate your data under the assumption of neutrality or selection under a relevant demographic scenario, then run the test on the simulated datasets to estimate the rates of false positives and false negatives given a threshold. A software like SliM3, msms or discoal can be used to perform these simulations. Comparing the results from different approaches as suggested earlier is also valuable.
Once you have your list of genomic windows, you might want to perform permutation tests to see how likely it is that you detect MYB by chance given that you have n windows classified as selected. The R package regioneR may be helpful for that.
Good luck!
I guess p-value shows you how much you can trust that this is a true event. You can also look around how to calculate the FDR (False Discovery Rate).
Also I wouldn't use wording such as "by chance" when talking about statistics. All data follow some kind of distribution. If for example the data follow a normal distribution, the bulk of the data points will be within 1 SD from the mean. The further away, the less density will be there. So it is not by chance that you observe or don't observe something, rather it is expected at such a level/count/whatever under this model distribution that you have selected for your data.
There is a nice new paper on selective sweeps:
Article How well do we understand the basis of classic selective swe...
DCMS method of combining results from multiple selection tests is nice for this purpose. See: Article Properties of different selection signature statistics and a...
Here is a successful example of exploiting this method:
Article Scans for signatures of selection in Russian cattle breed ge...
- Badges
- Science topic
- Similar topics
- Physics
- Condensed Matter Physics
- Solid State Physics
- PEAKS