I would like to ask you for your sugestions for selecting a test to detect signatures of selection in the following mouse model:

We have three groups: animals exhibiting trait A, trait B and controls. These animals were selected over the last 4 decades (controls were mated randomly and do not show any of the traits). That is total 170 generations (~4 generations per year).

We want to detect the genomic signatures of selection for trait A and B.

I am new in population genomics, but according to this paper (http://www.ncbi.nlm.nih.gov/pubmed/21218185), showing a desition tree (fig 1). I should apply a the Fst and a Linckage 'Desequilibrium test, given that the time scale would be short (40 years, 170 generations) and there are multiple populations.

Could you corroborate if this is the correct approach?


  • $\begingroup$ Do you assume you know what loci are affecting the trait? What kind of data do you have? There are a number of different approaches. Here, here and there are some papers that should help you. $\endgroup$
    – Remi.b
    Jan 5 '16 at 16:35
  • $\begingroup$ Thanks. I have no clue about the loci affecting the trait. The purpose of the study is to define that. We will do resequencing. $\endgroup$
    – Sergio.pv
    Jan 6 '16 at 7:52
  • $\begingroup$ Ok. You really have a number of different existing algorithm for that. I am not able here to make a good list and comparison of these algorithm but the aforementioned papers should help. One other set of question, you must ask yourself is "Do I know in which environment the individuals are selected in?", "Are the individuals occurring over a continuum of environmental gradient or are they occurring in either of two different environments?". $\endgroup$
    – Remi.b
    Jan 6 '16 at 15:48
  • $\begingroup$ I hope I got the question right: The mouse lines have been artificially selected in the same lab facility. From a founder population, animals were grouped by trait A or B. Within this groups, animals were mated over ~170 generation, avoiding inbreeding. A control line was kept, mated randomly and does not exhibit neither of the traits. $\endgroup$
    – Sergio.pv
    Jan 7 '16 at 8:19
  • $\begingroup$ Where is the selection process? Only the separation according to their trait value? $\endgroup$
    – Remi.b
    Jan 7 '16 at 14:56

I am assuming in this answer that you do not have any information about what environment is affecting the selection pressure so that methods like Bayenv2 can't be used.

The standard algorithms are:

  • Lewontin-Krakauer test
  • fdist
  • BayeScan
  • FLK
  • PCAdapt

Whitlock and Lotterhos 2014 showed that most often FLK and Bayenv2 perform better than the the three others. A number of articles (Meirmans 2012, Bierne et al 2013, De Mita et al 2013 and Fourcade et al 2013) have also shown that fdist and BayeScan suffer from high false positive rate. I would therefore recommend to go with FLK but I am probably not good enough to give very good advice.

Note that you should learn a bit about how these algorithms work and not using them blindly.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.