Simulation-Based Model Validation#
We used SLiM 3 to perform forward simulations
as a means to validate the diffusion approximation for collapsed polyploid spectra
in dadi. Below we give details on how to run each of the validation scripts.
The scripts can be found in the validation/ folder in the GitHub repo. To
run multiple replicates, each call to SLiM can be wrapped with a Bash for
loop with the loop index variable being passed to the rep parameter,
followed by any other parameters needed by the model and then the name of the
SLiM script:
for i in {1..100}
do
slim -d "rep=${i}" [other parameters] <script-name>
done
When run, each script generates a single realization of the demographic model
and outputs a site frequency spectrum that is the format required by dadi.
The corresponding Python code for comparing results from the diffusion approximation
in dadi can be found in the supp/SupplementalSfsData/plot_slim_dadi_comp.py
script.
SLiM scripts#
autotetraploid_snm.slim#
The standard neutral model for an autotetraploid.
# Parameters: rep
slim -d "rep=1" autotetraploid_snm.slim
autotetraploid_bottleneck.slim#
Simulates a bottleneck to population size nuBot times the original
population size.
# Parameters: rep, nuBot
slim -d "rep=1" -d "nuBot=0.5" autotetraploid_bottleneck.slim
allotetraploid_iso.slim#
Simulates an equilibrium population that splits at time T1 in the past
before being sampled.
# Parameters: rep, T1
slim -d "rep=1" -d "T1=0.5" allotetraploid_iso.slim
allotetraploid_bottleneck.slim#
Simulates an equilibrium population that splits at time T1 + T2 in the past
before being sampled, with a bottleneck of size nuBot occurring at time
T2 in the past.
# Parameters: rep, T1, T2, nuBot
slim -d "rep=1" -d "T1=0.5" -d "T2=0.25" -d "nuBot=0.5" allotetraploid_iso.slim
segtetraploid_iso.slim#
Simulates an equilibrium population that splits at time T1 + T2 in the past
before being sampled, with homoeologous exchange occurring at a rate of dij
starting at time T2 in the past. Here dij is equivalent to the exchange
parameter \(e_{i \leftrightarrow j}\) described in the manuscript. The use
of iso to describe this model is a bit of a misnomer since the subgenomes
are not isolated, but we kept it for naming consistency with the allotetraploid
models.
# Parameters: rep, T1, T2, dij
slim -d "rep=1" -d "T1=0.5" -d "T2=0.25" -d "dij=0.001" segtetraploid_iso.slim
segtetraploid_bottleneck.slim#
Simulates an equilibrium population that splits at time T1 + T2 in the past
before being sampled, with a bottleneck of size nuBot and homoeologous
exchange occurring at a rate of dij starting at time T2 in the past.
Here dij is equivalent to the exchange parameter
\(e_{i \leftrightarrow j}\) described in the manuscript.
# Parameters: rep, T1, T2, dij, nuBot
slim -d "rep=1" -d "T1=0.5" -d "T2=0.25" -d "dij=0.001" -d "nuBot=0.5" segtetraploid_iso.slim
Inference and plotting scripts#
In the inference/optimization/ folder, there are Python and R scripts to
simulate data under the allotetraploid and segmental allotetraploid models, run
inference on the simulated data with dadi, and then collate and plot the maximum
likelihood parameter estimates across all replicates for each combination of
parameters. The subfolders for each model contain the simulated frequency spectra
and optimization results (in CSV format) for all combinations of parameters tested.
References
Haller, B. C. and P. W. Messer. 2019. SLiM 3: Forward genetic simulations beyond the Wright–Fisher model. Molecular Biology and Evolution 36:632–-637.