def get_spectrum(self, location_batch):
"""TODO(nh2tran): docstring."""
#~ print("".join(["="] * 80)) # section-separating line
#~ print("WorkerIO.get_spectrum()")
spectrum_list = []
for location in location_batch:
# parse a spectrum
(precursor_mz,
charge,
scan,
raw_sequence,
mz_list,
intensity_list) = self._parse_spectrum(location)
# skip if precursor_mass > MZ_MAX
precursor_mass = precursor_mz * charge - deepnovo_config.mass_H * charge
if precursor_mass > self.MZ_MAX:
self.spectrum_count["skipped"] += 1
self.spectrum_count["skipped_mass"] += 1
continue
self.spectrum_count["read"] += 1
# pre-process spectrum
(spectrum_holder,
spectrum_original_forward,
spectrum_original_backward) = process_spectrum(mz_list,
intensity_list,
precursor_mass)
# update dataset
spectrum = {"scan": scan,
"precursor_mass": precursor_mass,
"spectrum_holder": spectrum_holder,
"spectrum_original_forward": spectrum_original_forward,
"spectrum_original_backward": spectrum_original_backward}
spectrum_list.append(spectrum)
return spectrum_list
def _parse_spectrum(self, precursor_mz, precursor_mass, rt_mean, scan_list,
ms1_list, input_file_handle):
"""TODO(nh2tran): docstring."""
#~ print("".join(["="] * 80)) # section-separating line
#~ print("WorkerIO: _parse_spectrum()")
spectrum_holder_list = []
spectrum_original_forward_list = []
spectrum_original_backward_list = []
### select best neighbors from the scan_list by their distance to rt_mean
# probably move this selection to get_location(), run once rather than repeating
neighbor_count = len(scan_list)
best_scan_index = None
best_distance = float('inf')
for scan_index, scan in enumerate(scan_list):
distance = abs(self.spectrum_rtinseconds_dict[scan] - rt_mean)
if distance < best_distance:
best_distance = distance
best_scan_index = scan_index
neighbor_center = best_scan_index
neighbor_left_count = neighbor_center
neighbor_right_count = neighbor_count - neighbor_left_count - 1
neighbor_size_half = self.neighbor_size // 2
neighbor_left_count = min(neighbor_left_count, neighbor_size_half)
neighbor_right_count = min(neighbor_right_count, neighbor_size_half)
### padding zero arrays to the left if not enough neighbor spectra
if neighbor_left_count < neighbor_size_half:
for x in range(neighbor_size_half - neighbor_left_count):
spectrum_holder_list.append(
np.zeros(shape=(1, self.MZ_SIZE), dtype=np.float32))
spectrum_original_forward_list.append(
np.zeros(shape=(1, self.MZ_SIZE), dtype=np.float32))
spectrum_original_backward_list.append(
np.zeros(shape=(1, self.MZ_SIZE), dtype=np.float32))
### parse and add neighbor spectra
scan_list_middle = []
ms1_intensity_list_middle = []
for index in range(neighbor_center - neighbor_left_count,
neighbor_center + neighbor_right_count + 1):
scan = scan_list[index]
scan_list_middle.append(scan)
ms1_entry = ms1_list[index]
ms1_intensity = float(re.split(':', ms1_entry)[1])
ms1_intensity_list_middle.append(ms1_intensity)
ms1_intensity_max = max(ms1_intensity_list_middle)
assert ms1_intensity_max > 0.0, "Error: Zero ms1_intensity_max"
ms1_intensity_list_middle = [
x / ms1_intensity_max for x in ms1_intensity_list_middle
]
for scan, ms1_intensity in zip(scan_list_middle,
ms1_intensity_list_middle):
spectrum_location = self.spectrum_location_dict[scan]
input_file_handle.seek(spectrum_location)
# parse header lines
line = input_file_handle.readline()
assert "BEGIN IONS" in line, "Error: wrong input BEGIN IONS"
line = input_file_handle.readline()
assert "TITLE=" in line, "Error: wrong input TITLE="
line = input_file_handle.readline()
assert "PEPMASS=" in line, "Error: wrong input PEPMASS="
line = input_file_handle.readline()
assert "CHARGE=" in line, "Error: wrong input CHARGE="
line = input_file_handle.readline()
assert "SCANS=" in line, "Error: wrong input SCANS="
line = input_file_handle.readline()
assert "RTINSECONDS=" in line, "Error: wrong input RTINSECONDS="
# parse fragment ions
mz_list, intensity_list = self._parse_spectrum_ion(
input_file_handle)
# pre-process spectrum
(spectrum_holder, spectrum_original_forward,
spectrum_original_backward) = process_spectrum(
mz_list, intensity_list, precursor_mass)
# normalize by each individual spectrum
#~ spectrum_holder /= np.max(spectrum_holder)
#~ spectrum_original_forward /= np.max(spectrum_original_forward)
#~ spectrum_original_backward /= np.max(spectrum_original_backward)
# weight by ms1 profile
#~ spectrum_holder *= ms1_intensity
#~ spectrum_original_forward *= ms1_intensity
#~ spectrum_original_backward *= ms1_intensity
# add spectrum to the neighbor list
spectrum_holder_list.append(spectrum_holder)
spectrum_original_forward_list.append(spectrum_original_forward)
spectrum_original_backward_list.append(spectrum_original_backward)
### padding zero arrays to the right if not enough neighbor spectra
if neighbor_right_count < neighbor_size_half:
for x in range(neighbor_size_half - neighbor_right_count):
spectrum_holder_list.append(
np.zeros(shape=(1, self.MZ_SIZE), dtype=np.float32))
spectrum_original_forward_list.append(
np.zeros(shape=(1, self.MZ_SIZE), dtype=np.float32))
spectrum_original_backward_list.append(
np.zeros(shape=(1, self.MZ_SIZE), dtype=np.float32))
spectrum_holder = np.vstack(spectrum_holder_list)
spectrum_original_forward = np.vstack(spectrum_original_forward_list)
spectrum_original_backward = np.vstack(spectrum_original_backward_list)
assert spectrum_holder.shape == (self.neighbor_size,
self.MZ_SIZE), "Error:shape"
# spectrum-CNN normalization: by feature
spectrum_holder /= np.max(spectrum_holder)
# ms1_profile
for x in range(neighbor_size_half - neighbor_left_count):
ms1_intensity_list_middle = [0.0] + ms1_intensity_list_middle
for x in range(neighbor_size_half - neighbor_right_count):
ms1_intensity_list_middle = ms1_intensity_list_middle + [0.0]
assert len(ms1_intensity_list_middle
) == self.neighbor_size, "Error: ms1 profile"
ms1_profile = np.array(ms1_intensity_list_middle)
return spectrum_holder, spectrum_original_forward, spectrum_original_backward, scan_list_middle, scan_list, ms1_profile