The biggest selling point of the slendr package is programming spatiotemporal population genetics models and executing them automatically by SLiM. However, there are several reasons why you might be interested in using slendr even for non-spatial models. First, R is a language that many scientist already know and being able to simulate data from the comfort of the R interface significantly lowers the barrier of entry. Second, because slendr makes SLiM appear almost as if it were just another R library, running simulations (spatial and non-spatial), fitting models, exploring parameter grids, calculating statistics and visualization of results can be all performed without leaving the R interface.
In this vignette, we will demonstrate how to program non-spatial models in slendr. However, we should start by noting that there is almost no difference between code for non-spatial and spatial models in slendr. The only visible difference is that the latter includes a
map = argument in the
population() constructor function of ancestral population(s), and the former does not. That’s it, that’s the difference. Switching between spatial and non-spatial models is performed internally by the package, without any user intervention.
To make the comparison clearer, we will use the example from the slendr landing page, but we will implement it in a non-spatial context (i.e. as a traditional random mating simulation).
First, lets define population objects, splits, and other demographic events (note the missing
map argument, which is set to
FALSE by default):
# African ancestral population afr <- population("AFR", time = 52000, N = 3000) # first migrants out of Africa ooa <- population("OOA", parent = afr, time = 51000, N = 500, remove = 25000) # Eastern hunter-gatherers ehg <- population("EHG", parent = ooa, time = 28000, N = 1000, remove = 6000) # European population eur <- population("EUR", parent = ehg, time = 25000, N = 2000) %>% resize(N = 10000, how = "exponential", time = 5000, end = 0) # Anatolian farmers ana <- population("ANA", time = 28000, N = 3000, parent = ooa, remove = 4000) # Yamnaya steppe population yam <- population("YAM", time = 7000, N = 500, parent = ehg, remove = 2500)
We can define geneflow events in the same way as we did for spatial models:
Compilation step is also the same. The only (internal) difference is that we skip the rasterization of vector maps that is performed for spatial models in order to control and restrict population boundaries:
plot() function doesn’t make sense, as there are no spatial maps to plot. However, we can still plot the demographic graph, verifying that the model has been specified correctly:
Even the final step—execution of the model in SLiM—is the same, using our builtin
slim(model, seq_length = 1, recomb_rate = 0, method = "batch")
Even for non-spatial models, this function still uses the same SLiM backend script used for spatial models. The only difference is that all spatial features are switched off, making the model run as a simple random-mating simulation.