Matthieu Foll 2011

Detecting natural selection from population-based genetic data

I would be very happy to help if you have some questions regarding BayeScan. However, before writing me, please ask yourself the following questions:
  • Did I read completely the manual?
  • My input file is correctly formatted? Even if you used software to create it, try to open it using a text editor. I personally recommend using Notepad++ or TextPad under windows, which are free and very powerful.
  • BayeScan is working as expected using the example input files provided? Compare your input file with the example files provided. Try to oversimplify your input file to find the problem (keep only one population and one locus for example).
  • If you are trying to use the console version, do you know the basic of commands? If not you can read this 10 minutes tutorial.
  • If you have difficulties to use the R functions, do you know the basics of R? If not you can read this introduction.
  • If you are trying to compile BayeScan for a specific platform, do you have a C++ compiler installed? If you don’t even know what this means, you can ask any informatician around you, even if he has no clue about population genetics… Also you will certainly have to remove the “-static” option in the Makefile on line 4 to compile under Mac OS.
  • If you are not sure whether BayeScan can be applied on your favorite species data set, I have to admit I certainly won’t be able to know this better than you. The basic assumptions behind BayeScan are summarized in the introduction of the manual, and you can find more details in the related publications. Each violation of an assumption can lead to an excess of false positives or eventually false negatives, but this is generally difficult to predict.  The correct way to do is to make simulations mimicking your real data and to see what BayeScan will produce. For example Fastsimcoal provides a very flexible way to simulate genetic data over a wide range of demographic scenario.
  • If you compare BayeScan with other methods aiming at identifying makers under selection (FDistDetSel etc.) and find different results, that’s maybe not a bug, but a nice feature! We are working hard to increase the power of our methods and to reduce false positives, so if the new methods keep giving the same answers as previous ones, this is useless.
Of course there are certainly some hidden bugs in BayeScan, and I would be grateful if you identify one of them… Despite the previous recommendations you should follow before contacting me, don’t hesitate to send me a message or to have a look at my web page: