def eic_centroid_detector(self, rt, eic, max_eic):
max_prominence = self.chromatogram_settings.peak_max_prominence_percent
max_height = self.chromatogram_settings.peak_height_max_percent
signal_threshold = self.chromatogram_settings.eic_signal_threshold
min_peak_datapoints = self.chromatogram_settings.min_peak_datapoints
peak_derivative_threshold = self.chromatogram_settings.peak_derivative_threshold
correct_baseline = False
include_indexes = sp.peak_picking_first_derivative(
rt,
eic,
max_height,
max_prominence,
max_eic,
min_peak_datapoints,
peak_derivative_threshold,
signal_threshold=signal_threshold,
correct_baseline=correct_baseline)
return include_indexes
def plot(self,
ax=None,
color="black",
derivative=True,
deriv_color='red'): #pragma: no cover
if self._ms_parent:
import matplotlib.pyplot as plt
if ax is None:
ax = plt.gca()
x = self._ms_parent.mz_exp_profile[self.start_scan:self.final_scan]
y = self._ms_parent.abundance_profile[self.start_scan:self.
final_scan]
ax.plot(x, y, color=color, label="Data")
ax.set(xlabel='m/z', ylabel='abundance')
if derivative and not self._ms_parent.is_centroid:
dy = sp.derivate(
self._ms_parent.
abundance_profile[self.start_index:self.final_index + 1])
ax.plot(x, dy, c=deriv_color)
else:
ax.plot((self.mz_exp, self.mz_exp), (0, self.abundance),
color=color,
label="Data")
#plt.legend()
return ax
else:
print("Centroid Peak Object")
def centroid_detector(self, rt, tic):
#noise_std = self.chromatogram_settings.std_noise_threshold
#method = self.chromatogram_settings.noise_threshold_method
#peak picking
#min_height = self.chromatogram_settings.peak_height_min_percent
min_peak_datapoints = self.chromatogram_settings.min_peak_datapoints
peak_derivative_threshold = self.chromatogram_settings.peak_derivative_threshold
signal_threshold = self.chromatogram_settings.peak_height_min_percent
# baseline detection
max_prominence = self.chromatogram_settings.peak_max_prominence_percent
max_height = self.chromatogram_settings.peak_height_max_percent
correct_baseline = False
#peak_indexes_generator = sp.peak_detector_generator(tic, noise_std, method, rt, max_height, min_height, max_prominence, min_datapoints)
include_indexes = sp.peak_picking_first_derivative(
rt,
tic,
max_height,
max_prominence,
max(tic),
min_peak_datapoints,
peak_derivative_threshold,
signal_threshold=signal_threshold,
correct_baseline=correct_baseline)
return include_indexes
def calc_centroid(self, mass, abund, freq):
max_height = self.mspeaks_settings.peak_height_max_percent
max_prominence = self.mspeaks_settings.peak_max_prominence_percent
min_peak_datapoints = self.mspeaks_settings.min_peak_datapoints
peak_derivative_threshold = self.mspeaks_settings.peak_derivative_threshold
max_abun = max(abund)
peak_height_diff = lambda hi, li: (
(abund[hi] - abund[li]) / max_abun) * 100
domain = mass
signal = abund
len_signal = len(signal)
signal_threshold, factor = self.get_threshold(abund)
max_signal = factor
correct_baseline = False
include_indexes = sp.peak_picking_first_derivative(
domain,
signal,
max_height,
max_prominence,
max_signal,
min_peak_datapoints,
peak_derivative_threshold,
signal_threshold=signal_threshold,
correct_baseline=correct_baseline,
abun_norm=1,
plot_res=False)
for indexes_tuple in include_indexes:
apex_index = indexes_tuple[1]
mz_exp_centroid, freq_centr, intes_centr = self.find_apex_fit_quadratic(
mass, abund, freq, apex_index)
if mz_exp_centroid:
peak_indexes = self.check_prominence(abund, apex_index,
len_signal,
peak_height_diff)
if peak_indexes:
peak_resolving_power = self.calculate_resolving_power(
abund, mass, apex_index)
s2n = intes_centr / self.baselise_noise_std
self.add_mspeak(self.polarity,
mz_exp_centroid,
abund[apex_index],
peak_resolving_power,
s2n,
indexes_tuple,
exp_freq=freq_centr,
ms_parent=self)
def smooth_signal(self, signal):
implemented_smooth_method = self.chromatogram_settings.implemented_smooth_method
pol_order = self.chromatogram_settings.savgol_pol_order
window_len = self.chromatogram_settings.smooth_window
window = self.chromatogram_settings.smooth_method
return sp.smooth_signal(signal, window_len, window, pol_order, implemented_smooth_method)
def smooth_signal(signal, parameters: LCMSParameters):
implemented_smooth_method = parameters.lc_ms.implemented_smooth_method
pol_order = parameters.lc_ms.savgol_pol_order
smooth_method = parameters.lc_ms.smooth_method
window_len = parameters.lc_ms.smooth_window
return sp.smooth_signal(signal, window_len, smooth_method, pol_order,
implemented_smooth_method)
def plot_detected_baseline(self, ax=None, color="blue"): # pragma: no cover
import matplotlib.pyplot as plt
if ax is None:
ax = plt.gca()
max_height = self.chromatogram_settings.peak_height_max_percent
max_prominence = self.chromatogram_settings.peak_max_prominence_percent
baseline = sp.baseline_detector(self.tic, self.retention_time, max_height, max_prominence)
ax.plot(self.retention_time, color=color)
ax.set(xlabel='Retention Time (s)', ylabel='Total Ion Chromatogram')
return ax
def centroid_detector(self, tic, rt):
''' this function has been replaced with sp.peak_picking_first_derivative
and it not used
'''
noise_std = self.chromatogram_settings.std_noise_threshold
method = self.chromatogram_settings.noise_threshold_method
''' peak picking'''
min_height = self.chromatogram_settings.peak_height_min_percent
min_datapoints = self.chromatogram_settings.min_peak_datapoints
''' baseline detection'''
max_prominence = self.chromatogram_settings.peak_max_prominence_percent
max_height = self.chromatogram_settings.peak_height_max_percent
peak_indexes_generator = sp.peak_detector_generator(tic, noise_std, method, rt, max_height, min_height, max_prominence, min_datapoints)
return peak_indexes_generator
def centroid_detector(self, tic, rt):
noise_std = self.chromatogram_settings.std_noise_threshold
method = self.chromatogram_settings.noise_threshold_method
#peak picking
min_height = self.chromatogram_settings.peak_height_min_percent
min_datapoints = self.chromatogram_settings.min_peak_datapoints
# baseline detection
max_prominence = self.chromatogram_settings.peak_max_prominence_percent
max_height = self.chromatogram_settings.peak_height_max_percent
peak_indexes_generator = sp.peak_detector_generator(
tic, noise_std, method, rt, max_height, min_height, max_prominence,
min_datapoints)
return peak_indexes_generator
def eic_centroid_detector(max_tic,
eic_data: EIC_Data,
parameters: LCMSParameters,
smooth=True):
# Do peak picking and store results inside EIC_Data class
max_prominence = parameters.lc_ms.peak_max_prominence_percent
max_height = parameters.lc_ms.peak_height_max_percent
signal_threshold = parameters.lc_ms.eic_signal_threshold
min_peak_datapoints = parameters.lc_ms.min_peak_datapoints
correct_baseline = parameters.lc_ms.correct_eic_baseline
peak_derivative_threshold = parameters.lc_ms.peak_derivative_threshold
if smooth:
eic_signal = smooth_signal(eic_data.eic, parameters)
else:
eic_signal = eic_data.eic
peak_indexes_generator = sp.peak_picking_first_derivative(
eic_data.time,
eic_signal,
max_height,
max_prominence,
max(eic_signal),
min_peak_datapoints,
peak_derivative_threshold,
signal_threshold=signal_threshold,
correct_baseline=correct_baseline,
check_abundance=True,
plot_res=False,
)
eic_data.apexes = [i for i in peak_indexes_generator]
def centroid_detector(self, tic, rt):
# need to change the parameter to accommodate EIC peak picking
# needs a better algorithm to detect start and end of a peak
noise_std = self.chromatogram_settings.std_noise_threshold
method = self.chromatogram_settings.noise_threshold_method
#peak picking
min_height = self.chromatogram_settings.peak_height_min_percent
min_datapoints = self.chromatogram_settings.min_peak_datapoints
# baseline detection
max_prominence = self.chromatogram_settings.peak_max_prominence_percent
max_height = self.chromatogram_settings.peak_height_max_percent
peak_indexes_generator = sp.peak_detector_generator(
tic, noise_std, method, rt, max_height, min_height, max_prominence,
min_datapoints)
return peak_indexes_generator
def deconvolution(self, peaks_entity_data, plot_res):
plot_res = True
domain = self.retention_time
signal = self._processed_tic
max_height = self.chromatogram_settings.peak_height_max_percent
max_prominence = self.chromatogram_settings.peak_max_prominence_percent
min_peak_datapoints = self.chromatogram_settings.min_peak_datapoints
signal_threshold = self.chromatogram_settings.peak_height_min_percent
max_rt_distance = self.chromatogram_settings.max_rt_distance
max_signal = max(signal)
correct_baseline = False
include_indexes = sp.peak_picking_first_derivative(domain, signal, max_height, max_prominence, max_signal, min_peak_datapoints,
signal_threshold=signal_threshold, correct_baseline=correct_baseline, plot_res=False)
''' deconvolution window is defined by the TIC peak region'''
all_apexes_rt = np.array(list(peaks_entity_data.keys()))
'''workaround for peak picking missing some local minimas'''
self.processed_appexes = []
for indexes_tuple in include_indexes:
start_rt = self.retention_time[indexes_tuple[0]]
#apex_rt = self.retention_time[indexes_tuple[1]]
final_rt = self.retention_time[indexes_tuple[2]]
''' find all features within TIC peak window'''
peak_features_indexes = np.where((all_apexes_rt > start_rt) & (all_apexes_rt < final_rt))[0]
peak_features_rts = all_apexes_rt[peak_features_indexes]
#print(start_rt, apex_rt, final_rt )
filtered_features_rt = []
filtered_features_abundance = []
for each_apex_rt in peak_features_rts:
apex_data = peaks_entity_data.get(each_apex_rt).get(each_apex_rt)
peak_features_tic = sum(peaks_entity_data.get(each_apex_rt).get(each_apex_rt).get('abundance'))
norm_smooth_tic = (peak_features_tic/ max_signal)*100
''' TODO:
Improve Peak Filtering
Calculate peaks sharpness here and filter it out (Amax - An /n)?
Peak Fit and Calculate Peak Gaussian Similarity?
Currentely using flat % tic relative abundance threshold and min 3 m/z per mass spectrum
'''
if norm_smooth_tic > signal_threshold and len(apex_data['mz']) > 1:
#print(len(apex_data['mz']))
filtered_features_rt.append(each_apex_rt)
filtered_features_abundance.append(peak_features_tic)
if len(filtered_features_rt) > 1:
''' more than one peak feature identified inside a TIC peak '''
# plt.plot(self.retention_time[indexes_tuple[0]:indexes_tuple[2]], signal[indexes_tuple[0]:indexes_tuple[2]], c='black')
#print(filtered_features_rt)
grouped_rt = self.hc(filtered_features_rt, filtered_features_abundance, max_rt_distance=max_rt_distance)
#print(grouped_rt)
for group, apex_rt_list in grouped_rt.items():
''' each group is a peak feature defined by the hierarchical clutter algorithm
'''
group_datadict = {}
group_datadict['ref_apex_rt'] = []
for each_group_apex_rt in apex_rt_list:
datadict = peaks_entity_data.get(each_group_apex_rt)
for rt, each_datadict in datadict.items():
if rt == "ref_apex_rt":
group_datadict['ref_apex_rt'].append(each_datadict)
else:
if rt in group_datadict.keys():
mz_list = each_datadict.get("mz")
abundance_list = each_datadict.get("abundance")
each_mz_abun = dict(zip(mz_list, abundance_list))
for index_mz, mz in enumerate(group_datadict[rt].get("mz")):
if mz in each_mz_abun.keys():
each_mz_abun[mz] = each_mz_abun[mz] + group_datadict[rt].get("abundance")[index_mz]
else:
each_mz_abun[mz] = group_datadict[rt].get("abundance")[index_mz]
group_datadict[rt] = { 'mz': list(each_mz_abun.keys()) ,
'abundance': list(each_mz_abun.values()),
'scan_number': each_datadict.get('scan_number') }
else:
group_datadict[rt] = each_datadict
peak_rt = []
peak_tic = []
#print(group_datadict.get('ref_apex_rt'))
for rt, each_datadict in group_datadict.items():
if rt != "ref_apex_rt":
peak_rt.append(rt)
peak_tic.append(sum(each_datadict["abundance"]))
peak_rt, peak_tic = zip(*sorted(zip(peak_rt, peak_tic)))
smoothed_tic = self.smooth_signal(peak_tic)
include_indexes = sp.peak_picking_first_derivative(peak_rt, smoothed_tic, max_height, max_prominence, max_signal, min_peak_datapoints,
signal_threshold=signal_threshold, correct_baseline=False, plot_res=False)
include_indexes = list(include_indexes)
if include_indexes:
if len(include_indexes) > 1:
''' after sum there are two apexes
check if it is inside the deconvolution window, otherwise ignores it
'''
for new_apex_index in include_indexes:
# pass
self.add_gcpeak(new_apex_index, start_rt, final_rt, peak_rt, smoothed_tic, group_datadict, plot_res)
else:
''' after sum there is on apex
save it
'''
new_apex_index = include_indexes[0]
#print(include_indexes, group, apex_rt_list)
self.add_gcpeak(new_apex_index, start_rt, final_rt, peak_rt, smoothed_tic, group_datadict, plot_res)
elif len(filtered_features_rt) == 1:
''' only one peak feature inside deconvolution window '''
each_apex_rt = filtered_features_rt[0]
datadict = peaks_entity_data.get(each_apex_rt)
peak_rt = []
peak_tic = []
for rt, each_datadict in datadict.items():
if rt != "ref_apex_rt":
peak_rt.append(rt)
peak_tic.append(sum(each_datadict["abundance"]))
peak_rt, peak_tic = zip(*sorted(zip(peak_rt, peak_tic)))
smoothed_tic = self.smooth_signal(peak_tic)
include_indexes = sp.peak_picking_first_derivative(peak_rt, smoothed_tic, max_height, max_prominence, max_signal, min_peak_datapoints,
signal_threshold=signal_threshold, correct_baseline=False, plot_res=False)
include_indexes = list(include_indexes)
if include_indexes:
''' after sum there are two apexes
check if it is inside the deconvolution window, otherwise ignores it'''
if len(include_indexes) > 1:
for new_apex_index in include_indexes:
# pass
self.add_gcpeak(new_apex_index, start_rt, final_rt, peak_rt, smoothed_tic, datadict, plot_res)
else:
''' after sum there is one apex
save it
includes_indexes = (start, apex, final )'''
new_apex_index = include_indexes[0]
self.add_gcpeak(new_apex_index, start_rt, final_rt, peak_rt, smoothed_tic, datadict, plot_res)
else:
# print('no data after filter')
pass
if plot_res:
plt.plot(self.retention_time, self._processed_tic, c='black')
plt.show()
def find_peaks_entity(self, eic_dict):
''' combine eic with mathing rt apexes'''
max_prominence = self.chromatogram_settings.peak_max_prominence_percent
max_height = self.chromatogram_settings.peak_height_max_percent
signal_threshold = self.chromatogram_settings.eic_signal_threshold
min_peak_datapoints = self.chromatogram_settings.min_peak_datapoints
correct_baseline = False
peaks_entity_data = {}
max_eic = 0
for mz, eic_scan_index_rt in eic_dict.items():
ind_max_eic = max(eic_scan_index_rt[0])
max_eic = ind_max_eic if ind_max_eic > max_eic else max_eic
for mz, eic_scan_index_rt in eic_dict.items():
eic = eic_scan_index_rt[0]
rt_list = eic_scan_index_rt[1]
if len(eic) >= min_peak_datapoints:
smooth_eic = self.smooth_tic(eic)
include_indexes = sp.peak_picking_first_derivative(rt_list, smooth_eic, max_height, max_prominence, max_eic, min_peak_datapoints,
signal_threshold=signal_threshold, correct_baseline=correct_baseline)
for initial_scan, apex_scan, final_scan in include_indexes:
rt_corrected_therm = self.quadratic_interpolation(rt_list, smooth_eic, apex_scan)
ref_apex_rt = round(rt_list[apex_scan] + rt_corrected_therm,4)
apex_rt = rt_list[apex_scan]
# apex_abundance = smooth_eic[apex_scan]
#maximum_tic = apex_abundance if apex_abundance > maximum_tic else maximum_tic
for scan_index in range(initial_scan, final_scan):
peak_rt = rt_list[scan_index]
peak_abundance = smooth_eic[scan_index]
dict_data = {peak_rt: { 'mz': [mz] ,
'abundance':[peak_abundance],
'scan_number': [scan_index] },
"ref_apex_rt": ref_apex_rt
}
if not apex_rt in peaks_entity_data.keys():
peaks_entity_data[apex_rt] = dict_data
else:
if not peak_rt in peaks_entity_data[apex_rt].keys():
peaks_entity_data[apex_rt][peak_rt] = dict_data.get(peak_rt)
else:
existing_data = peaks_entity_data[apex_rt].get(peak_rt)
existing_data['mz'].append(mz)
existing_data['abundance'].append(peak_abundance)
existing_data['scan_number'].append(scan_index)
return peaks_entity_data
