def _split_data(datalist, withhold):
"""Splits a list of IfnData instances into test and train subsets, placing
<withold> percentage of the data in the test subset. The testing subset is
then aligned using a DataAlignment instance, and the training subset is
scaled according to the *testing* subset scale factors. The test and train
aligned IfnData objects output by the DataAlignment.summarize_data() method
are returned.
"""
assert 0 <= withhold <= 100
# Build mask which selects <withhold> points for test subset
data_coord = _get_data_coordinates(datalist[0])
test_size = int((100-withhold) * len(data_coord) / 100.0)
test_idcs = np.random.choice(len(data_coord), test_size, False)
test_coord = [data_coord[i] for i in test_idcs]
train_coord = [c for c in data_coord if c not in test_coord]
# Separate data into test and train subsets
test_datalist = [d.copy() for d in datalist]
train_datalist = [d.copy() for d in datalist]
for obj in test_datalist:
for c in test_coord:
obj.data_set.loc[c[0:2]][c[2]] = np.NaN
for obj in train_datalist:
for c in train_coord:
obj.data_set.loc[c[0:2]][c[2]] = np.NaN
train_alignment = DataAlignment()
train_alignment.add_data(train_datalist)
train_alignment.align()
train_alignment.get_scaled_data()
train = train_alignment.summarize_data()
if withhold == 0:
test = None
else:
test_alignment = DataAlignment()
test_alignment.add_data(test_datalist)
test_alignment.scale_factors = train_alignment.scale_factors
test_alignment.get_scaled_data()
test = test_alignment.summarize_data()
return train, test
def timecourse_figure():
# Get experimental data
newdata_1 = IfnData("20190108_pSTAT1_IFN_Bcell")
newdata_2 = IfnData("20190119_pSTAT1_IFN_Bcell")
newdata_3 = IfnData("20190121_pSTAT1_IFN_Bcell")
newdata_4 = IfnData("20190214_pSTAT1_IFN_Bcell")
# Aligned data, to get scale factors for each data set
alignment = DataAlignment()
alignment.add_data([newdata_4, newdata_3, newdata_2, newdata_1])
alignment.align()
alignment.get_scaled_data()
mean_data = alignment.summarize_data()
# Plot
green = sns.color_palette("deep")[2]
red = sns.color_palette("deep")[3]
light_green = sns.color_palette("pastel")[2]
light_red = sns.color_palette("pastel")[3]
plot = TimecoursePlot((1, 1))
plot.add_trajectory(mean_data,
'errorbar',
'o--', (0, 0),
label=r'10 pM IFN$\alpha$2',
color=light_red,
dose_species='Alpha',
doseslice=10.0,
alpha=0.5)
plot.add_trajectory(mean_data,
'errorbar',
'o--', (0, 0),
label=r'6 pM IFN$\beta$',
color=light_green,
dose_species='Beta',
doseslice=6.0,
alpha=0.5)
plot.add_trajectory(mean_data,
'errorbar',
'o-', (0, 0),
label=r'3000 pM IFN$\alpha$2',
color=red,
dose_species='Alpha',
doseslice=3000.0)
plot.add_trajectory(mean_data,
'errorbar',
'o-', (0, 0),
label=r'2000 pM IFN$\beta$',
color=green,
dose_species='Beta',
doseslice=2000.0)
fname = os.path.join(os.getcwd(), 'results', 'Figures', 'Figure_4',
'Timecourse.pdf')
plot.axes.set_ylabel('pSTAT1 (MFI)')
plot.show_figure(show_flag=False, save_flag=True, save_dir=fname)
from ifnclass.ifnfit import StepwiseFit
if __name__ == '__main__':
# ------------------------------
# Align all data
# ------------------------------
newdata_1 = IfnData("20190108_pSTAT1_IFN_Bcell")
newdata_2 = IfnData("20190119_pSTAT1_IFN_Bcell")
newdata_3 = IfnData("20190121_pSTAT1_IFN_Bcell")
newdata_4 = IfnData("20190214_pSTAT1_IFN_Bcell")
alignment = DataAlignment()
alignment.add_data([newdata_4, newdata_3, newdata_2, newdata_1])
alignment.align()
alignment.get_scaled_data()
mean_data = alignment.summarize_data()
# -------------------------------
# Initialize model
# -------------------------------
times = newdata_4.get_times('Alpha')
doses_alpha = newdata_4.get_doses('Alpha')
doses_beta = newdata_4.get_doses('Beta')
Mixed_Model = IfnModel('Mixed_IFN_ppCompatible')
Mixed_Model.set_parameters({
'R2': 4920,
'R1': 1200,
'k_a1': 2.0e-13,
'k_a2': 1.328e-12,
'k_d3': 1.13e-4,
'k_d4': 0.9,
# | 1E-10 6E-12
# G| alpha beta
# | 1E-11 2E-13
# H| alpha beta
# | 0 pM 0 pM
# Load saved DataAlignment
small_alignment = DataAlignment()
small_alignment.load_from_save_file(
'small_alignment', os.path.join(os.getcwd(), 'small_alignment'))
large_alignment = DataAlignment()
large_alignment.load_from_save_file(
'large_alignment', os.path.join(os.getcwd(), 'large_alignment'))
small_alignment.align()
small_alignment.get_scaled_data()
mean_small_data = small_alignment.summarize_data()
large_alignment.align()
large_alignment.get_scaled_data()
mean_large_data = large_alignment.summarize_data()
# ----------------------
# Set up Figure layout
# ----------------------
# Set up dose response figures
new_fit = DoseresponsePlot((1, 2), figsize=(9, 1.5 * 2.5))
new_fit.axes[0].set_ylabel('pSTAT1 (MFI)')
# new_fit.axes[1].set_ylabel('pSTAT1 (MFI)')
# Plot Dose respsonse data
times = [2.5, 5.0, 7.5, 10.0, 20.0, 60.0]
color_palette = sns.color_palette("rocket_r",
def score_params(params):
# --------------------
# Set up Model
# --------------------
Mixed_Model, DR_method = lm.load_model()
scale_factor, DR_KWARGS = lm.SCALE_FACTOR, lm.DR_KWARGS
Mixed_Model.set_parameters(params)
# Make predictions
times = [2.5, 5.0, 7.5, 10.0, 20.0, 60.0]
alpha_doses = [10, 100, 300, 1000, 3000, 10000, 100000]
beta_doses = [0.2, 6, 20, 60, 200, 600, 2000]
dra60 = DR_method(times,
'TotalpSTAT',
'Ia',
alpha_doses,
parameters={'Ib': 0},
sf=scale_factor,
**DR_KWARGS)
drb60 = DR_method(times,
'TotalpSTAT',
'Ib',
beta_doses,
parameters={'Ia': 0},
sf=scale_factor,
**DR_KWARGS)
sim_df = IfnData('custom',
df=pd.concat((dra60.data_set, drb60.data_set)),
conditions={
'Alpha': {
'Ib': 0
},
'Beta': {
'Ia': 0
}
})
# --------------------
# Set up Data
# --------------------
newdata_1 = IfnData("20190108_pSTAT1_IFN_Bcell")
newdata_2 = IfnData("20190119_pSTAT1_IFN_Bcell")
newdata_3 = IfnData("20190121_pSTAT1_IFN_Bcell")
newdata_4 = IfnData("20190214_pSTAT1_IFN_Bcell")
# Aligned data, to get scale factors for each data set
alignment = DataAlignment()
alignment.add_data([newdata_4, newdata_3, newdata_2, newdata_1])
alignment.align()
alignment.get_scaled_data()
mean_data = alignment.summarize_data()
# --------------------
# Score model
# --------------------
sim_df.drop_sigmas()
mean_data.drop_sigmas()
# rmse = RMSE(mean_data.data_set.values, sim_df.data_set.values)
mae = MAE(mean_data.data_set.values, sim_df.data_set.values)
return mae