phi = Integration.one_pop(phi, xx, T188, nu188)
phi = Integration.one_pop(phi, xx, T189, nu189)
phi = Integration.one_pop(phi, xx, T190, nu190)
phi = Integration.one_pop(phi, xx, T191, nu191)
phi = Integration.one_pop(phi, xx, T192, nu192)
phi = Integration.one_pop(phi, xx, T193, nu193)
phi = Integration.one_pop(phi, xx, T194, nu194)
phi = Integration.one_pop(phi, xx, T195, nu195)
phi = Integration.one_pop(phi, xx, T196, nu196)
phi = Integration.one_pop(phi, xx, T197, nu197)
phi = Integration.one_pop(phi, xx, T198, nu198)
# simulate sfs:
fs = Spectrum.from_phi(phi,ns,(xx,))
return fs
SMCPLUS_extrap_function_198 = Numerics.make_extrap_log_func(SMCPLUS_model_198steps)
#################################### CEU ##################################
# get SFS relative to theta = 1
smc_song_fs_ceu_relT = SMCPLUS_extrap_function_198(song_ceu_dadi,ns,pts)
# get the proportional sfs:
smc_song_fs_ceu_freq = smc_song_fs_ceu_relT / np.sum(smc_song_fs_ceu_relT)
# Plotting.plot_1d_comp_Poisson(obs_fs_CEU/np.sum(obs_fs_CEU),smc_song_fs_ceu_freq)
# get num snp, need a theta:
# Length is from Gutenkunst so that SFSes are comparable
Length = 4040000
# wsong mu: 1.25 × 10−8
song_mu= 1.25e-08
# ancestral size (oldest nu)
song_ceu_nuA = song_ceu_dips[-1]
phi = Integration.one_pop(phi, xx, T30, nu30)
phi = Integration.one_pop(phi, xx, T31, nu31)
phi = Integration.one_pop(phi, xx, T32, nu32)
phi = Integration.one_pop(phi, xx, T33, nu33)
phi = Integration.one_pop(phi, xx, T34, nu34)
phi = Integration.one_pop(phi, xx, T35, nu35)
phi = Integration.one_pop(phi, xx, T36, nu36)
phi = Integration.one_pop(phi, xx, T37, nu37)
phi = Integration.one_pop(phi, xx, T38, nu38)
# get expected SFS:
fs = Spectrum.from_phi(phi, ns, (xx, ))
return fs
# wrap your function in Numerics.make_extrap_log_func to make extrapolation version of your function
extrap_nso_model_trim_39_function = Numerics.make_extrap_log_func(
nso_model_trim_39)
# Version of model with nu and T values plugged in plus one more epoch for dadi to optimize
def nso_model_trim_39_plusContraction_forOptimization(
(contractionGen_RescBy2Nanc, contractionSize_RescByNanc), ns, pts):
xx = Numerics.default_grid(pts)
# intialize phi with ancestral pop: (nu0 = 1)
phi = PhiManip.phi_1D(xx, nu=1)
# stays at nu0=1 for T0 duration of time:
# followed by a number of time steps, with associated pop changes:
phi = Integration.one_pop(phi, xx, 0.012177548731, 1)
phi = Integration.one_pop(phi, xx, 0.00712323213699992, 0.444441187582237)
phi = Integration.one_pop(phi, xx, 0.00505418013100008, 0.444441187582237)
phi = Integration.one_pop(phi, xx, 0.00392030906399995, 0.137194521186969)
phi = Integration.one_pop(phi, xx, 0.00320305953600002, 0.137194521186969)
#pop_ids=["O1"]
#projections=[32]
import os
# replace this with your appropriate dir name
# os.chdir("/Users/c-monstr/Documents/allRAD_august2015/digitifera/dadi")
dd = Misc.make_data_dict(infile)
data = Spectrum.from_data_dict(dd, pop_ids, projections, polarized=True)
ns = data.sample_sizes
pts = [65, 80, 95]
np.set_printoptions(precision=3)
#-------------------
gr = Numerics.make_extrap_log_func(Demographics1D.growth)
params = array([3, 1])
upper_bound = [100, 100]
lower_bound = [0.01, 0.01]
poptg = dadi.Inference.optimize_log(params,
data,
gr,
pts,
lower_bound=lower_bound,
upper_bound=upper_bound,
verbose=len(params),
maxiter=15)
model = gr(poptg, ns, pts)
phi = Integration.one_pop(phi, xx, T14, nu14)
phi = Integration.one_pop(phi, xx, T15, nu15)
phi = Integration.one_pop(phi, xx, T16, nu16)
phi = Integration.one_pop(phi, xx, T17, nu17)
phi = Integration.one_pop(phi, xx, T18, nu18)
phi = Integration.one_pop(phi, xx, T19, nu19)
# get expected SFS:
fs = Spectrum.from_phi(phi,ns,(xx,))
return fs
func=nso_model_trim_20
modelName="nso_model_trim_20"
params=params # change for each model
# wrap your function in Numerics.make_extrap_log_func to make extrapolation version of your function
func_ex = Numerics.make_extrap_log_func(func)
model=func_ex(params,ns,pts_l) # this is relative to theta =1
### get proportional sfs
model_freq = model/(sum(model))
model_freq_fold =model_freq.fold()
############### Output SFS ########################
print('Writing out SFS **************************************************')
outputSFS=str(outdir)+"/"+str(modelName)+".PROPORTIONAL.FOLDED.expSFS.txt"
model_freq_fold.to_file(outputSFS)
phi = Integration.one_pop(phi, xx, T13, nu13)
phi = Integration.one_pop(phi, xx, T14, nu14)
phi = Integration.one_pop(phi, xx, T15, nu15)
phi = Integration.one_pop(phi, xx, T16, nu16)
phi = Integration.one_pop(phi, xx, T17, nu17)
phi = Integration.one_pop(phi, xx, T18, nu18)
phi = Integration.one_pop(phi, xx, T19, nu19)
phi = Integration.one_pop(phi, xx, T20, nu20)
phi = Integration.one_pop(phi, xx, T21, nu21)
# get expected SFS:
fs = Spectrum.from_phi(phi, ns, (xx, ))
return fs
# wrap your function in Numerics.make_extrap_log_func to make extrapolation version of your function
extrap_sso_model_trim_22_function = Numerics.make_extrap_log_func(
sso_model_trim_22)
# Version of model with nu and T values plugged in plus one more epoch for dadi to optimize
def sso_model_trim_22_plusContraction_forOptimization(
(contractionGen_RescBy2Nanc, contractionSize_RescByNanc), ns, pts):
xx = Numerics.default_grid(pts)
# intialize phi with ancestral pop: (nu0 = 1)
phi = PhiManip.phi_1D(xx, nu=1)
# stays at nu0=1 for T0 duration of time:
# followed by a number of time steps, with associated pop changes:
phi = Integration.one_pop(phi, xx, 0.0871326536999987, 1)
phi = Integration.one_pop(phi, xx, 0.0826612257000005, 1)
phi = Integration.one_pop(phi, xx, 0.0786283415999992, 1.01729865557432)
phi = Integration.one_pop(phi, xx, 0.0749695096499999, 1.01729865557432)
phi = Integration.one_pop(phi, xx, 0.0716360027500006, 1.07723241130487)
phi = Integration.one_pop(phi, xx, T30, nu30)
phi = Integration.one_pop(phi, xx, T31, nu31)
phi = Integration.one_pop(phi, xx, T32, nu32)
phi = Integration.one_pop(phi, xx, T33, nu33)
phi = Integration.one_pop(phi, xx, T34, nu34)
phi = Integration.one_pop(phi, xx, T35, nu35)
phi = Integration.one_pop(phi, xx, T36, nu36)
phi = Integration.one_pop(phi, xx, T37, nu37)
phi = Integration.one_pop(phi, xx, T38, nu38)
# get sfs:
fs = Spectrum.from_phi(phi, ns, (xx, ))
return fs
# wrap your function in Numerics.make_extrap_log_func to extrapolate over pts
msmc_extrap_function_38 = Numerics.make_extrap_log_func(msmc_model_38steps)
# usage: msmc_extrap_function_38(Parameters,ns,pts)
# where Parameters is a vector of MSMC output parameters
# converted to population sizes across time intervals, in units of 2*nuA (see below)
###### dadi model for 8 Haplotype MSMC model (28 steps) #########################
# need 28 steps for 8 haplotype model
def msmc_model_28steps(
(nu28, nu27, nu26, nu25, nu24, nu23, nu22, nu21, nu20, nu19, nu18, nu17,
nu16, nu15, nu14, nu13, nu12, nu11, nu10, nu9, nu8, nu7, nu6, nu5, nu4,
nu3, nu2, nu1, nuA, T28, T27, T26, T25, T24, T23, T22, T21, T20, T19, T18,
T17, T16, T15, T14, T13, T12, T11, T10, T9, T8, T7, T6, T5, T4, T3, T2,
T1, T0), ns, pts):
xx = Numerics.default_grid(pts)
np.set_printoptions(precision=3)
#-------------------
def two_growths((nu1, nu2, T1, T2), (n1, ), pts):
xx = Numerics.default_grid(pts)
phi = PhiManip.phi_1D(xx)
nu_func1 = lambda t: np.exp(np.log(nu1) * t / T1)
phi = Integration.one_pop(phi, xx, T1, nu_func1)
nu_func2 = lambda t: nu1 * np.exp(np.log(nu2 / nu1) * t / T2)
phi = Integration.one_pop(phi, xx, T2, nu_func2)
sfs = Spectrum.from_phi(phi, (n1, ), (xx, ))
return sfs
twog = Numerics.make_extrap_log_func(two_growths)
params = array([1, 1, 1, 0.1])
upper_bound = [100, 100, 100, 10]
lower_bound = [0.01, 0.01, 0.01, 0.001]
poptg = dadi.Inference.optimize_log(params,
data,
twog,
pts,
lower_bound=lower_bound,
upper_bound=upper_bound,
verbose=len(params),
maxiter=20)
model = twog(poptg, ns, pts)
#-------------------
# split with migration
def s2mM(params, ns, pts):
nu1, nu2, T, M12, M21 = params
xx = Numerics.default_grid(pts)
phi = PhiManip.phi_1D(xx)
phi = PhiManip.phi_1D_to_2D(xx, phi)
phi = Integration.two_pops(phi, xx, T, nu1, nu2, m12=M12, m21=M21)
fs = Spectrum.from_phi(phi, ns, (xx, xx))
return fs
S2MM = Numerics.make_extrap_log_func(s2mM)
params = array([1, 1, 1, 30, 30])
upper_bound = [10, 10, 10, 300, 300]
lower_bound = [0.01, 0.01, 0.01, 0.1, 0.1]
poptg = dadi.Inference.optimize_log(params,
data,
S2MM,
pts,
lower_bound=lower_bound,
upper_bound=upper_bound,
verbose=len(params),
maxiter=15)
model = S2MM(poptg, ns, pts)
print 's2m ', sys.argv[1], sys.argv[2], sys.argv[3], ' params : ', poptg
nu1_func = lambda t: nu01 * (nu1 / nu01)**(t / T)
nu2_func = lambda t: nu02 * (nu2 / nu02)**(t / T)
phi = Integration.two_pops(phi,
xx,
T,
nu1_func,
nu2_func,
m12=m12,
m21=m21)
fs = Spectrum.from_phi(phi, ns, (xx, xx))
return fs
im2 = Numerics.make_extrap_log_func(IM2)
params = array([0.078, 2.136, 2.174, 2.063, 1.174, 0.514, 0.379])
upper_bound = [100, 100, 100, 100, 100, 200, 200]
lower_bound = [1e-3, 1e-3, 1e-3, 1e-3, 1e-3, 0, 0]
poptg = dadi.Inference.optimize_log(params,
data,
im2,
pts,
lower_bound=lower_bound,
upper_bound=upper_bound,
verbose=len(params),
maxiter=20)
model = im2(poptg, ns, pts)
ll_model = dadi.Inference.ll_multinom(model, data)
phi = Integration.one_pop(phi, xx, T24, nu24) phi = Integration.one_pop(phi, xx, T25, nu25) phi = Integration.one_pop(phi, xx, T26, nu26) phi = Integration.one_pop(phi, xx, T27, nu27) phi = Integration.one_pop(phi, xx, T28, nu28) phi = Integration.one_pop(phi, xx, T29, nu29) phi = Integration.one_pop(phi, xx, T30, nu30) phi = Integration.one_pop(phi, xx, T31, nu31) phi = Integration.one_pop(phi, xx, T32, nu32) # get expected SFS: fs = Spectrum.from_phi(phi,ns,(xx,)) return fs # wrap your function in Numerics.make_extrap_log_func to make extrapolation version of your function extrap_sso_model_trim_33_function = Numerics.make_extrap_log_func(sso_model_trim_33) msmc_model_expSFS=extrap_sso_model_trim_33_function(params,ns,pts_l) ll_msmc_model = dadi.Inference.ll_multinom(msmc_model_expSFS, fs) # plot an image: #import pylab import matplotlib.pyplot as plt #fig=plt.figure(1) #pylab.ion() outputFigure=str(str(outdir)+"/"+str(pop)+"."+str(modelName)+".wDip.figure.png") dadi.Plotting.plot_1d_comp_multinom(msmc_model_expSFS, fs) pyplot.title(str(modelName)+" LL = "+str(ll_msmc_model))
# split with migration, with recent change in migration
def s2mRM(params, ns, pts):
nu1, nu2, T, M12, M21, M12r, M21r = params
Tr = 0.01
xx = Numerics.default_grid(pts)
phi = PhiManip.phi_1D(xx)
phi = PhiManip.phi_1D_to_2D(xx, phi)
phi = Integration.two_pops(phi, xx, T, nu1, nu2, m12=M12, m21=M21)
phi = Integration.two_pops(phi, xx, Tr, nu1, nu2, m12=M12r, m21=M21r)
fs = Spectrum.from_phi(phi, ns, (xx, xx))
return fs
S2MRM = Numerics.make_extrap_log_func(s2mRM)
params = array([1, 1, 1, 30, 30, 30, 30])
upper_bound = [10, 10, 10, 300, 300, 300, 300]
lower_bound = [0.01, 0.01, 0.01, 0.1, 0.1, 0.1, 0.1]
poptg = dadi.Inference.optimize_log(params,
data,
S2MRM,
pts,
lower_bound=lower_bound,
upper_bound=upper_bound,
verbose=len(params),
maxiter=15)
model = S2MRM(poptg, ns, pts)
print 's2mRM ', sys.argv[1], sys.argv[2], sys.argv[3], ' params : ', poptg
