def test_musyc_no_gamma():
import numpy as np
from synergy.combination import MuSyC
from synergy.utils.dose_tools import grid
E0, E1, E2, E3 = 1, 0.2, 0.1, 0
h1, h2 = 2.3, 0.8
C1, C2 = 1e-2, 1e-1
alpha12, alpha21 = 3.2, 1.1
model = MuSyC(E0=E0,
E1=E1,
E2=E2,
E3=E3,
h1=h1,
h2=h2,
C1=C1,
C2=C2,
alpha12=alpha12,
alpha21=alpha21,
variant="no_gamma")
npoints = 8
npoints2 = 12
D1, D2 = grid(1e-3, 1e0, 1e-2, 1e1, npoints, npoints2)
np.random.seed(0)
E = model.E(D1, D2)
Efit = E * (1 + (np.random.rand(len(D1)) - 0.5) / 10.)
model.fit(D1, D2, Efit)
assert model.r_squared > 0.9
def test_musyc_higher():
import numpy as np
from synergy.utils import dose_tools
from synergy.higher import MuSyC
from synergy.utils import plots
E_params = [2, 1, 1, 1, 1, 0, 0, 0]
h_params = [2, 1, 0.8]
C_params = [0.1, 0.01, 0.1]
alpha_params = [2, 3, 1, 1, 0.7, 0.5, 2, 1, 1]
gamma_params = [0.4, 2, 1, 2, 0.7, 3, 2, 0.5, 2]
params = E_params + h_params + C_params + alpha_params + gamma_params
truemodel = MuSyC()
truemodel.parameters = params
d = dose_tools.grid_multi((1e-3, 1e-3, 1e-3), (1, 1, 1), (6, 6, 6),
include_zero=True)
d1 = d[:, 0]
d2 = d[:, 1]
d3 = d[:, 2]
np.random.seed(0)
E = truemodel.E(d)
noise = 0.05
E_fit = E + noise * (E_params[0] -
E_params[-1]) * (2 * np.random.rand(len(E)) - 1)
model = MuSyC(E_bounds=(0, 2),
h_bounds=(1e-3, 1e3),
alpha_bounds=(1e-5, 1e5),
gamma_bounds=(1e-5, 1e5))
model.fit(d, E_fit)
assert model.r_squared > 0.9
npoints,
npoints2,
replicates=replicates,
include_zero=True)
D1 = D1 / 3.
E = model.E(D1, D2)
Efit = E * (1 + (np.random.rand(len(D1)) - 0.5) / 5.)
scatter_points = pd.DataFrame({
'drug1.conc': D1,
'drug2.conc': D2,
'effect': Efit
})
model.fit(D1, D2, Efit)
D1, D2 = dose_tools.grid(1e-3,
1,
1e-2,
10,
npoints * 10,
npoints2 * 10,
include_zero=True)
D1 = D1 / 3
model.plot_surface_plotly(D1,
D2,
scatter_points=scatter_points,
xlabel="Drug1",
ylabel="Drug2",
npoints2 = 8
d1, d2 = grid(C1 / 1e2,
C1 * 1e2,
C2 / 1e2,
C2 * 1e2,
npoints,
npoints2,
include_zero=True)
E = truemodel.E(d1, d2)
noise = 0.05
E_fit = E + noise * (E0 - E3) * (2 * np.random.rand(len(E)) - 1)
model = MuSyC()
model.fit(d1, d2, E_fit, bootstrap_iterations=100)
#print(model)
#if model.converged:
# print(model.get_parameter_range().T)
print(model.summary())
if False:
scatter_points = pd.DataFrame({
'drug1.conc': d1,
'drug2.conc': d2,
'effect': E_fit
})
DD1, DD2 = grid(C1 / 1e4, C1 * 1e4, C2 / 1e4, C2 * 1e4, npoints * 2,
npoints2 * 2)
schindler = Schindler()
s_synergy = schindler.fit(d1,
d2,
E,
drug1_model=loewe.drug1_model,
drug2_model=loewe.drug2_model)
bliss = Bliss()
bsynergy = bliss.fit(d1,
d2,
E,
drug1_model=loewe.drug1_model,
drug2_model=loewe.drug2_model)
musyc = MuSyC()
musyc.fit(d1, d2, E)
fig = plt.figure(figsize=(12, 3))
ax = fig.add_subplot(1, 4, 1)
musyc.plot_heatmap(d1, d2, ax=ax, logscale=False, title="Sham Surface")
ax = fig.add_subplot(1, 4, 2)
loewe.plot_heatmap(ax=ax,
title="-log(Loewe)",
vmin=-0.01,
vmax=0.01,
logscale=False,
neglog=True)
ax.set_yticks([])