def nbd_fret_err(nbd_data, fret_data, plot=False, plot_title=None):
"""Utility function for calculating NBD and FRET data error."""
# Plot titles
if plot_title:
nbd_plot_title = plot_title + ' NBD'
fret_plot_title = plot_title + ' Bid/Bax FRET'
else:
nbd_plot_title = None
fret_plot_title = None
# Calculate NBD err
nbd_residuals, _ = calc_err_var(nbd_data, last_n_pts=80, plot=plot,
plot_title=nbd_plot_title)
nbd_err = np.std(nbd_residuals, ddof=1)
# Calculate FRET err
fret_residuals, _ = calc_err_var(fret_data, last_n_pts=80, plot=plot,
plot_title=fret_plot_title)
fret_err = np.std(fret_residuals, ddof=1)
return (nbd_err, fret_err)
nbd_residue, rep_num, 'TIME')].values
# The fitting procedure expects a three-dimensional array for
# each entry; the first dimension is reserved for multiple
# concentrations, as in a titration, which we don't have; the
# second is reserved for multiple observables, and the third
# contains the timepoints.
data_var = np.zeros((1, len(observables), len(time_var)))
# Iterate over the observables
for obs_ix, obs in enumerate(observables):
timecourse = \
df[(activator, obs, nbd_residue, rep_num, 'VALUE')].values
data_var[0, obs_ix, :] = timecourse
# Get the residuals for this observable/rep
(residuals, fig) = \
calc_err_var(timecourse, last_n_pts=num_residuals,
fit_type='cubic', plot=False)
# Add the residuals for this rep to the list
data_residuals[0, obs_ix, :] = residuals
# NBD-specific variables
if obs == 'NBD':
# Initial fluorescence value
f0 = timecourse[0]
# Update maximum fluorescence value
max_nbd_for_rep = np.max(timecourse)
if max_nbd_for_rep > max_nbd_value:
max_nbd_value = max_nbd_for_rep
# Get the minimum fluorescence allowable for this rep:
# 0.1 * the initial value
lbound = f0 * 0.1
# Get the maximum allowable fluorescence for this rep: the
def find_outliers(values):
(res, fig) = calc_err_var(values, last_n_pts=len(values),
fit_type='two_exp_sum', plot=False)
zscores = mod_zscore(res)
return filtered(values, zscores, 3.5)
bgsub_wells = subtract_background_set(nbd_wells, bax_bg_wells)
# Background subtracted, averaged
#(bgsub_averages, bgsub_stds) = averages(bgsub_wells, layout)
# First timepoint shifted to 0 (better for fitting)
#reset_bgsub_means = reset_first_timepoint_to_zero(bgsub_norm_averages)
#Timecourses normalized, BG-subtracted, averaged, then with first point
#shifted to t = 0.
#reset_bgsub_sds = reset_first_timepoint_to_zero(bgsub_norm_stds)
# Get the time vector
time = bgsub_wells['Bax 0 nM, NBD-Bax 96 nM'][TIME]
# Initialize numpy data matrix
data_to_fit = np.zeros((len(bgsub_wells.keys()), 1, len(time)))
# Initialize matrix of experimental error values
data_sigma = np.zeros((len(bgsub_wells.keys()), 1))
conc_index = 0
# Normalize data by the initial values
bax_concs_to_fit = []
for conc_name in bgsub_wells.keys():
bax_concs_to_fit.append(float(conc_name.split(' ')[1]))
y = bgsub_wells[conc_name][VALUE]
y = y / np.mean(y[0:2])
data_to_fit[conc_index, 0, :] = y
(residuals, fig) = calc_err_var(y, last_n_pts=200, fit_type='quadratic',
plot=False)
data_sigma[conc_index, 0] = np.std(residuals, ddof=1)
conc_index += 1
nbd_bax_conc = 96.
if max_nbd_for_rep > max_nbd_value:
max_nbd_value = max_nbd_for_rep
# Get the minimum fluorescence allowable for this rep: 0.1 * the
# initial value
lbound = f0 * 0.1
# Get the maximum allowable fluorescence for this rep: the lesser
# of 10 times the initial value or a corrected absolute NBD
# fluorescence of 300k
max_nbd_fluorescence = 300000
mutant_absolute_max = \
labeling_ratios[nbd_residue] * max_nbd_fluorescence
mutant_relative_max = 10 * f0
ubound = min(mutant_absolute_max, mutant_relative_max)
# Get the residuals for this rep
(residuals, fig) = \
calc_err_var(timecourse, last_n_pts=50, fit_type='cubic',
plot=False)
# Add the residuals for this rep to the list
residuals_reps.append(residuals)
# Variable names
time_var_name = 'time_%s_%s_%s_r%s' % \
(activator, obs, nbd_residue, rep_num)
data_var_name = 'data_%s_%s_%s_r%s' % \
(activator, obs, nbd_residue, rep_num)
# Upper and lower bounds for fluorescence parameters
lbound_name = data_var_name + '_lbound'
ubound_name = data_var_name + '_ubound'
f0_name = data_var_name + '_f0'
# Add these as module-level variables
setattr(this_module, time_var_name, time_var)
setattr(this_module, data_var_name, data_var)
setattr(this_module, lbound_name, lbound)
# Matrices to use in fitting
time_54 = df_pre[('Bid', 'FRET', '54', 1, 'TIME')].values
data_54 = np.zeros((1, 2, len(time_54)))
data_54[0, 0, :] = df_pre[('Bid', 'NBD', '54', 1, 'VALUE')].values
data_54[0, 1, :] = df_pre[('Bid', 'FRET', '54', 1, 'VALUE')].values
time_126 = df_pre[('Bid', 'FRET', '126', 1, 'TIME')].values
data_126 = np.zeros((1, 2, len(time_126)))
data_126[0, 0, :] = df_pre[('Bid', 'NBD', '126', 1, 'VALUE')].values
data_126[0, 1, :] = df_pre[('Bid', 'FRET', '126', 1, 'VALUE')].values
# Calculate the prior on standard error of the data by running polynomial
# fits on the final points
# 54
data_54_sigma = np.zeros((1, 2))
(residuals, _) = calc_err_var(data_54[0, 0, :], last_n_pts=80,
fit_type='cubic', plot=False)
data_54_sigma[0, 0] = np.std(residuals, ddof=1)
(residuals, _) = calc_err_var(data_54[0, 1, :], last_n_pts=80,
fit_type='cubic', plot=False)
data_54_sigma[0, 1] = np.std(residuals, ddof=1)
# 126
data_126_sigma = np.zeros((1, 2))
(residuals, _) = calc_err_var(data_126[0, 0, :], last_n_pts=80,
fit_type='cubic', plot=False)
data_126_sigma[0, 0] = np.std(residuals, ddof=1)
(residuals, _) = calc_err_var(data_126[0, 1, :], last_n_pts=80,
fit_type='cubic', plot=False)
data_126_sigma[0, 1] = np.std(residuals, ddof=1)
