def test_figs(data, gp):
xpred = np.zeros((100,2))
xpred[:50,0] = np.linspace(data.index.min(),data.index.max())
xpred[50:,0] = np.linspace(data.index.min(),data.index.max())
xpred[50:,1] = 1
mu,cov = gp.predict(xpred,full_cov=True)
var = np.diag(cov)
mu = mu[:,0]
fig = plt.figure()
plt.plot(xpred[:50,0],mu[:50],label=parent);
plt.fill_between(xpred[:50,0],mu[:50]-2*np.sqrt(var[:50]),mu[:50]+2*np.sqrt(var[:50]),alpha=.1)
plt.plot(xpred[:50,0],mu[50:],label='L2B5-500')
plt.fill_between(xpred[:50,0],mu[50:]-2*np.sqrt(var[50:]),mu[50:]+2*np.sqrt(var[50:]),alpha=.1)
plt.legend(fontsize=20)
fig_name = tt.get_path() + '/figures/uMax/test_gpr.png'
fig.savefig(fig_name, bbox_inches = "tight", pad_inches = 0.4, dpi = 600)
plt.close()
op = np.zeros((50,100))
op[np.arange(50),np.arange(50)] = -1
op[np.arange(50),np.arange(50)+50] = 1
oddeltaMu = np.dot(op,mu)
oddeltaCov = np.dot(np.dot(op,cov),op.T)
oddeltaVar = np.diag(oddeltaCov)
fig = plt.figure()
plt.plot([data.index.min(),data.index.max()],[0,0],'k',lw=3)
plt.plot(xpred[:50,0],oddeltaMu,label='od-delta')
plt.fill_between(xpred[:50,0],oddeltaMu-2*np.sqrt(oddeltaVar),oddeltaMu+2*np.sqrt(oddeltaVar),alpha=.1)
plt.legend(fontsize=20)
fig_name = tt.get_path() + '/figures/uMax/test_od_delta.png'
fig.savefig(fig_name, bbox_inches = "tight", pad_inches = 0.4, dpi = 600)
plt.close()
def plot_data(clean_data_path):
out_path = tt.get_path() + '/figures/uMax/data_plots/' + re.split(
'[./]', clean_data_path)[-2]
pathlib.Path(out_path).mkdir(exist_ok=True)
df = pd.read_csv(clean_data_path, sep='\t')
time_hours = get_time(df)
for column in df:
if ('blank' in column) or ('Time' in column) or ('Temp_C' in column):
continue
od = df[column].values
fig = plt.figure()
plt.scatter(time_hours,
od,
c='#87CEEB',
marker='o',
label='_nolegend_',
s=60)
plt.title(column, fontsize=24)
plt.xlabel('Time (hours)', fontsize=18)
plt.ylabel('OD 600', fontsize=18)
fig_name = out_path + '/' + column + '.png'
fig.savefig(fig_name, bbox_inches="tight",
pad_inches=0.4) #, dpi = 600)
plt.close()
def get_params(file_name):
df_path = tt.get_path(
) + '/data/uMax/clean_data/Task2_48hr_48well_discon_180614_140037.txt'
df = pd.read_csv(df_path, sep='\t')
df.index = tt.get_time(df)
data = tt.remove_unnecessary_columns(df)
data = data.loc[:,
data.columns.str.contains('|'.
join(['L0B2-500', 'L2B5-500']))]
data = data.loc[:, ~data.columns.str.
contains('|'.join(['L0B2-500_1_1', 'L0B2-500_3_2']))]
strains = [x.split('_')[0] for x in data.columns.values]
bio_reps = [x.split('_')[1] for x in data.columns.values]
tech_reps = [x.split('_')[2] for x in data.columns.values]
condition = ['control' for x in data.columns.values]
meta = pd.DataFrame({
'strain': strains,
'Bio': bio_reps,
'Tech': tech_reps,
'Condition': condition
})
#parent = 'L0B2-500'
parent = strains[0]
control = 'control'
#mutants = ['L2B5-500']
meta['strain-regression'] = (meta.strain != parent).astype(int)
meta['condition'] = (meta.Condition != control).astype(int)
meta['interaction'] = meta['strain-regression'] * meta.condition
bgreat.setGlobals(_data=data, _meta=meta)
bgreat.setGlobals(_parent=parent, _control=control)
bgreat.condition = None
meta.rename(columns={'strain': 'Strain'}, inplace=True)
data = pd.concat([meta, data.transpose()], axis=1)
# add index as well column for now
data['well'] = data.index.values
cols = list(data)
cols.insert(0, cols.pop(cols.index('well')))
data = data.ix[:, cols]
gpFactory = factory.Factory()
mse = metric.MeanSquareError(factory=gpFactory)
muMax = metric.MuMax_simple(factory=gpFactory, n=100)
#lagTime = metric.LagTime(f=gpFactory)
lagTime = metric.LagTime(
f=.5,
factory=gpFactory,
)
carryingCapacity = metric.CarryingCapacity_simple(factory=gpFactory)
output = pd.DataFrame()
timeind = 8
time = data.columns[timeind:].astype(float)
for i in range(output.shape[0], data.shape[0]):
od = np.log2(data.iloc[i, timeind:].tolist())
od = od - od[0]
design = params = data.iloc[i, :timeind]
params = pd.DataFrame([params.values],
columns=params.index,
index=[params.name])
key = pd.DataFrame(design).T
temp = GrowthData(
pd.DataFrame(list(zip(time.tolist(), od.tolist())),
columns=['time', params.index[0]]), key)
edata = temp.getData("gp")
n = edata.shape[0]
train = list(range(n))
test = list(range(0, n, 5))
[train.remove(x) for x in test]
cv = [train, test]
edata_train = edata.iloc[train, :]
edata_test = edata.iloc[test, :]
# GP MSE
try:
start = timemodule.time()
gp = gpFactory.build(edata_train, optimize=True)
stop = timemodule.time()
params['gp_mse'] = mse.compute(edata_test, gp)
params['gp_muMax_mean'], params['gp_muMax_std'] = muMax.compute(
predictive_data=temp.data.values, model=gp)
params['gp_CarryingCapacity_mean'], params[
'gp_CarryingCapacity_std'] = carryingCapacity.compute(
predictive_data=temp.data.values, model=gp)
#params['gp_LagTime_mean'],params['gp_LagTime_std'] = lagTime.compute(predictive_data=temp.data.values,model=gp)
#params['gp_LagTime_mean'] = lagTime.compute(gpFactory.buildInputFixed(time_min=0,time_max=max(edata.time),size=200,convert=False),gp)
#print(lagTime.compute(predictive_data = test_data, model = gp1))
params['gp_loglikelihood'] = gp.log_likelihood()
params['gp_time'] = stop - start
except Exception as e:
params['gp_mse'] = [np.nan]
params['gp_muMax_mean'], params['gp_muMax_std'] = [np.nan, np.nan]
params['gp_CarryingCapacity_mean'], params[
'gp_CarryingCapacity_std'] = [np.nan, np.nan]
#params['gp_LagTime_mean'],params['gp_LagTime_std'] = [np.nan,np.nan]
params['gp_loglikelihood'] = [np.nan]
params['gp_time'] = [np.nan]
if output.shape[0] == 0:
output = params
else:
output = output.append(params)
df_out = tt.get_path(
) + '/data/uMax/params_gpr/Task2_48hr_48well_discon_180614_140037.txt'
output.to_csv(df_out, sep='\t', index=False)
pathlib.Path(out_path).mkdir(exist_ok=True)
df = pd.read_csv(clean_data_path, sep='\t')
time_hours = get_time(df)
for column in df:
if ('blank' in column) or ('Time' in column) or ('Temp_C' in column):
continue
od = df[column].values
fig = plt.figure()
plt.scatter(time_hours,
od,
c='#87CEEB',
marker='o',
label='_nolegend_',
s=60)
plt.title(column, fontsize=24)
plt.xlabel('Time (hours)', fontsize=18)
plt.ylabel('OD 600', fontsize=18)
fig_name = out_path + '/' + column + '.png'
fig.savefig(fig_name, bbox_inches="tight",
pad_inches=0.4) #, dpi = 600)
plt.close()
in_path = tt.get_path(
) + '/data/uMax/raw_data/Task2_48hr_48well_discon_180614_140037/Task2_48hr_48well_discon_180614_140037.txt'
out_path = tt.get_path(
) + '/data/uMax/clean_data/Task2_48hr_48well_discon_180614_140037.txt'
#cleanData(in_path, out_path)
plot_data(out_path)
def get_params():
df_path = tt.get_path() + '/data/uMax/merged_clean_data/metadata.txt'
df = pd.read_csv(df_path, sep = '\t')