Files for the article "Demogenomic inference from spatially and temporally heterogeneous samples", Marchi, Kapopoulou, and Excoffier


##CODES##
The zipped directories linux.zip and win.zip contain the fastsimcoal2 linux and windows versions used for the analyses



##SIMULATION##
Simulated data for models with spatial or temporal heterogeneities and attempts to recover the parameters of these true scenarios with fastsimcoal2
3 type of heterogeneity have been tested:
	1) "Spatial2pop" = Spatial heterogeneity with 2 populations per continent (Fig 1A)
	2) "10PopDivPool" = Spatial heterogeneity with 5 populations per continent (Fig 1B)
	3) "Temporal" = Temporal Heterogeneity (Fig 1C-D)


	###SIMULATED DATA###
	Genetic data have been simulation for the different conditions (named def by fastsimcoal2) described in Table 1:

	1) "Spatial2pop" = Spatial heterogeneity with 2 populations per continent (Fig 1A) 
		- Def 1: Strong heterogeneity; No admixture
		- Def 2: Weak heterogeneity; No admixture
		- Def 3: No heterogeneity; No admixture
		- Def 4: Medium heterogeneity; Large admixture
		- Def 5: Medium heterogeneity; Small admixture
		- Def 6: Medium heterogeneity; No admixture

	2) "10PopDivPool" = Spatial heterogeneity with 5 populations per continent (Fig 1B)
		- Def 3: no admixture for n=2 
		- Def 6: no admixture for n=10 

	3) "Temporal" = Temporal Heterogeneity
		- Def 1: 2 populations per continent (Fig 1C)
		- Def 5: 1 population per continent (Fig 1D)

	10 replicated sfs have been simulated for each condition using the following command line: 
	./fsc -t xxx.tpl -f xxx.def -n10 -I -s0 -d -q -x -j -k1000000
		where xxx is an arbitrary generic name for the model; the conditions are given with the -f parameter; 
		the other parameters are described in fastsimcoal2 manual (http://cmpg.unibe.ch/software/fastsimcoal2/man/fastsimcoal27.pdf)


	###ESTIMATES###
	Attempt to recover true (=simulated) parameters with different strategies described in Fig 2:
		- "noStruct" = No Structure
		- "fis" = Implicit Structure
		- "pool" = Explicit Structure
		- "onepop" = Temporal Structure

	Parameter estimation was done with the following command line:
	./fsc -t xxx_i.tpl -e xxx_i.est -n500000 -M -c1 -B1 -d -L40 -q --logprecision 18 
		where xxx is a different generic name for each combination of simulation model and estimation strategy; i iterates 	over the 10 data sets.

	Likelihoods of the different tested strategies are shown in Lhood_xxx.txt in each model folder
	Parameters estimated for the different tested strategies are shown in Param_xxx.txt in each model folder


###############################

##SGDP##
Fastsimcoal input files for the 4 models tested on the SGDP data.

	###ESTIMATES###
	For each model, we estimated parameters with the following command line:
		./fsc -t xxx.tpl -e xxx.est -n200000 -L30 -q  -C1 -M -d --multiSFS --logprecision 18 -c1 -B1
			where xxx stands for the specific model to be studied.
	
	###CONFIDENCE INTERVALS###
	Confidence intervals for the parameters estimated under the best models were obtained using a parametric bootstrap approach.
	We first generated 100 SFS using the maximum-likelihood estimated parameter values:
		./fsc -i xxx_boot.par -n100 -j -d -s0 -x –I -q -u
		

	We then re-estimated the parameters of the model using 20 independent runs and starting around the estimated ML parameter values:
		./fsc -t xxx_boot.tpl -e xxx_boot.est –initvalues xxx_boot.pv -M -L30  -n500000 -d --multiSFS -C5 --logprecision 18 -q