4

My question is closely related to this post.

There are a number of existing platforms to perform numerical individual-based simulation in populations genetics. An almost exhaustive list of such platforms can be found here.

Some program (like NEMO or SFS_CODE for example) perform forward simulations while some others (like SIMCOAL2 for example) perform backward (coalescence) simulations.

I am asking for a general comparison between those two methods.

  • Is one faster than the other? Why? (Intuitively I'd expect backward simulations to be faster).
  • Does backward simulation allow for simulating selection, recombination, gender-specific mutation rate, complex demography, etc..
  • Why/when would one use a backward/forward simulation model rather than the other type of model?
Community
  • 1
Remi.b
  • 68,088
  • 11
  • 141
  • 234

1 Answers1

2

Your intuitions all seem correct. Coalescent simulation should be faster, because you don't track the entire population history over all t generations as you do in the forward simulation. Rather, as you work backwards in time you are tracking a smaller and small population. And with coalescent simulations it is probably very hard to incorporate the full range of evolutionary processes whereas doing so is more or less trivial with forward simulations.

A 2008 paper by Carvajal-Rodriguez reviews some of the issues in more detail that I have here.

Corvus
  • 1,781
  • 12
  • 20