def main(data_dir='/data'):
data_dir = pathlib.Path(data_dir)
if not data_dir.exists():
raise FileNotFoundError(f'Path not found!! {data_dir.as_posix()}')
# ==================== DEFINE CONSTANTS =====================
trial_type_mapper = {'CorrR': ('hit', 'right', 'no early'),
'CorrL': ('hit', 'left', 'no early'),
'ErrR': ('miss', 'right', 'no early'),
'ErrL': ('miss', 'left', 'no early'),
'NoLickR': ('ignore', 'right', 'no early'),
'NoLickL': ('ignore', 'left', 'no early'),
'LickEarly': ('non-performing', 'non-performing', 'early')
}
cell_type_mapper = {'PT': 'PT', 'IT': 'IT', 'unidentified': 'N/A', 'in': 'interneuron', 'pn': 'Pyr'}
task_protocol = {'task': 'audio delay', 'task_protocol': 1}
kargs = dict(skip_duplicates=True,
ignore_extra_fields=True,
allow_direct_insert=True)
# ================== INGESTION OF DATA ==================
data_files = data_dir.glob('*.nwb')
for data_file in data_files:
print(f'-- Read {data_file} --')
fname = data_file.stem + '.nwb'
nwb = h5py.File(os.path.join('/data', fname), 'r')
experimenter = nwb['general/experimenter'][()]
lab.Person.insert1((experimenter, experimenter), **kargs)
# ----------- meta-data -----------
search_result = re.search(
'nwb_(an\d+)_(\d+)_(.*)_(\d+)um.nwb', fname)
subject_nickname = search_result.group(1)
session_date = parse_date(search_result.group(2)).date()
brain_location = 'left_' + search_result.group(3).lower()
depth = int(search_result.group(4))
subject = dict(subject_nickname=subject_nickname,
username=experimenter,
sex='M', species='Mus musculus',
animal_source='Jackson Labs') # inferred from paper
modified_gene = dict(
gene_modification='Thy1-GCaMP6s',
gene_modification_description='GP4.3')
lab.ModifiedGene.insert1(modified_gene, **kargs)
lab.Subject.insert1(subject, skip_duplicates=True)
session_date = parse_date(
re.search(f'nwb_{subject_nickname}_(\d+)_', fname).group(1))
if len(experiment.Session & subject):
session = (lab.Subject & subject).aggr(
experiment.Session.proj(session_user='******'),
session_max='max(session)').fetch1('session_max') + 1
else:
session = 1
current_session = dict(
**subject, session_date=session_date,
brain_location_name=brain_location,
session=session)
if imaging.TrialTrace & current_session:
print('Data ingested, skipping over...')
continue
experiment.Session.insert1(current_session, **kargs)
experiment.Session.ImagingDepth.insert1(
dict(**current_session, imaging_depth=depth), **kargs)
# ---- trial data ----
print('---- Ingesting trial data ----')
(session_trials, behavioral_trials, trial_events) = [], [], []
pole_in = nwb['stimulus/presentation/pole_in/timestamps'][()]
pole_out = nwb['stimulus/presentation/pole_out/timestamps'][()]
auditory_cue = nwb['stimulus/presentation/auditory_cue/timestamps'][()]
for i_trial in range(len(nwb['epochs'])):
trial_num = i_trial + 1
trial_path = 'epochs/trial_{0:03}/'.format(trial_num)
tkey = dict(**current_session, trial=trial_num)
trial_start = nwb[trial_path + 'start_time'][()]
trial_stop = nwb[trial_path + 'stop_time'][()]
sample_start = pole_in[i_trial] - trial_start
delay_start = pole_out[i_trial] - trial_start
response_start = auditory_cue[i_trial] - trial_start
session_trials.append(
dict(**tkey, start_time=Decimal(trial_start),
stop_time=Decimal(trial_stop)))
outcome, trial_instruction, early = trial_type_mapper[nwb[trial_path + 'tags'][()][0]]
behavioral_trials.append(
dict(**tkey, **task_protocol,
trial_instruction=trial_instruction,
outcome=outcome,
early_lick=early))
for etype, etime in zip(('sample', 'delay', 'go'), (sample_start, delay_start, response_start)):
if not np.isnan(etime):
trial_events.append(
dict(**tkey, trial_event_id=len(trial_events)+1,
trial_event_type=etype, trial_event_time=etime))
# insert trial info
experiment.SessionTrial.insert(session_trials, **kargs)
experiment.BehaviorTrial.insert(behavioral_trials, **kargs)
experiment.TrialEvent.insert(trial_events, **kargs)
# ---- Scan info ----
print('---- Ingesting imaging data ----')
imaging.Scan.insert1(
dict(**current_session,
image_gcamp=nwb['acquisition/images/green'][()],
frame_time=nwb['processing/ROIs/ROI_001/timestamps'][()]),
**kargs)
tr_events = {tr: (float(stime), float(gotime)) for tr, stime, gotime in
zip(*(experiment.SessionTrial * experiment.TrialEvent
& current_session & 'trial_event_type = "go"').fetch(
'trial', 'start_time', 'trial_event_time'))}
rois, trial_traces = [], []
frame_time = nwb['processing/ROIs/ROI_001/timestamps'][()]
for i_roi in range(len(nwb['processing/ROIs'])):
roi_idx = i_roi + 1
roi_path = 'processing/ROIs/ROI_{0:03}/'.format(roi_idx)
roi_trace = nwb[roi_path + 'fmean'][()]
neuropil_trace = nwb[roi_path + 'fmean_neuropil'][()]
roi_trace_corrected = roi_trace - 0.7*neuropil_trace
rois.append(
dict(**current_session,
roi_idx=roi_idx,
roi_trace=roi_trace,
neuropil_trace=neuropil_trace,
roi_trace_corrected=roi_trace_corrected,
cell_type=cell_type_mapper[nwb[roi_path + 'cell_type'][()][0]],
ap_position=nwb[roi_path + 'AP_position_from_bregma_in_micron'][()][0],
ml_position=nwb[roi_path + 'ML_position_from_bregma_in_micron'][()][0],
roi_pixel_list=nwb[roi_path + 'pixel_list'][()],
inc=bool(np.mean(roi_trace)/np.mean(neuropil_trace)>1.05)))
for tr in tr_events.keys():
if tr in tr_events:
go_cue_time = sum(tr_events[tr])
go_id = np.abs(frame_time-go_cue_time).argmin()
idx = slice(go_id-70, go_id+45, 1)
baseline = np.mean(roi_trace_corrected[go_id-70:go_id-64])
trial_traces += [
dict(**current_session, roi_idx=roi_idx, trial=tr,
original_time=frame_time[idx],
aligned_time=frame_time[idx]-go_cue_time,
aligned_trace=roi_trace[idx],
aligned_trace_corrected=roi_trace_corrected[idx],
dff=(roi_trace_corrected[idx] - baseline)/baseline)]
imaging.Scan.Roi.insert(rois, **kargs)
imaging.TrialTrace.insert(trial_traces, **kargs)
def main(meta_data_dir='./data/meta_data', reingest=False):
meta_data_dir = pathlib.Path(meta_data_dir)
if not meta_data_dir.exists():
raise FileNotFoundError(f'Path not found!! {meta_data_dir.as_posix()}')
# ==================== DEFINE CONSTANTS =====================
# ---- inferred from paper ----
hemi = 'left'
skull_reference = 'bregma'
photostim_devices = {
473: 'LaserGem473',
594: 'LaserCoboltMambo100',
596: 'LaserCoboltMambo100'
}
# ---- from lookup ----
probe = 'A4x8-5mm-100-200-177'
electrode_config_name = 'silicon32'
# ================== INGESTION OF METADATA ==================
# ---- delete all Sessions ----
if reingest:
experiment.Session.delete()
# ---- insert metadata ----
meta_data_files = meta_data_dir.glob('*.mat')
for meta_data_file in tqdm(meta_data_files):
print(f'-- Read {meta_data_file} --')
meta_data = sio.loadmat(meta_data_file,
struct_as_record=False,
squeeze_me=True)['meta_data']
# ==================== person ====================
person_key = dict(username=meta_data.experimenters,
fullname=meta_data.experimenters)
lab.Person.insert1(person_key, skip_duplicates=True)
# ==================== subject gene modification ====================
modified_genes = (meta_data.animalGeneModification if isinstance(
meta_data.animalGeneModification,
(np.ndarray, list)) else [meta_data.animalGeneModification])
lab.ModifiedGene.insert(
(dict(gene_modification=g, gene_modification_description=g)
for g in modified_genes),
skip_duplicates=True)
# ==================== subject strain ====================
animal_strains = (meta_data.animalStrain if isinstance(
meta_data.animalStrain,
(np.ndarray, list)) else [meta_data.animalStrain])
lab.AnimalStrain.insert(zip(animal_strains), skip_duplicates=True)
# ==================== subject ====================
animal_id = (meta_data.animalID[0] if isinstance(
meta_data.animalID, (np.ndarray, list)) else meta_data.animalID)
animal_source = (meta_data.animalSource[0] if isinstance(
meta_data.animalSource,
(np.ndarray, list)) else meta_data.animalSource)
subject_key = dict(
subject_id=int(re.search('\d+', animal_id).group()),
sex=meta_data.sex[0].upper() if len(meta_data.sex) != 0 else 'U',
species=meta_data.species,
animal_source=animal_source)
try:
date_of_birth = parse_date(meta_data.dateOfBirth)
subject_key['date_of_birth'] = date_of_birth
except:
pass
lab.AnimalSource.insert1((animal_source, ), skip_duplicates=True)
with lab.Subject.connection.transaction:
if subject_key not in lab.Subject.proj():
lab.Subject.insert1(subject_key)
lab.Subject.GeneModification.insert(
(dict(subject_key, gene_modification=g)
for g in modified_genes),
ignore_extra_fields=True)
lab.Subject.Strain.insert(
(dict(subject_key, animal_strain=strain)
for strain in animal_strains),
ignore_extra_fields=True)
# ==================== session ====================
session_key = dict(
subject_key,
username=person_key['username'],
session=len(experiment.Session & subject_key) + 1,
session_date=parse_date(meta_data.dateOfExperiment + ' ' +
meta_data.timeOfExperiment))
experiment.Session.insert1(session_key, ignore_extra_fields=True)
print(
f'\tInsert Session - {session_key["subject_id"]} - {session_key["session_date"]}'
)
# ==================== Probe Insertion ====================
brain_location_key = (experiment.BrainLocation & dict(
brain_area=meta_data.extracellular.recordingLocation,
hemisphere=hemi,
skull_reference=skull_reference)).fetch1('KEY')
insertion_loc_key = dict(
brain_location_key,
ml_location=meta_data.extracellular.recordingCoordinates[0] *
1000, # mm to um
ap_location=meta_data.extracellular.recordingCoordinates[1] *
1000, # mm to um
dv_location=meta_data.extracellular.recordingCoordinates[2]
) # already in um
with ephys.ProbeInsertion.connection.transaction:
ephys.ProbeInsertion.insert1(dict(
session_key,
insertion_number=1,
probe=probe,
electrode_config_name=electrode_config_name),
ignore_extra_fields=True)
ephys.ProbeInsertion.InsertionLocation.insert1(
dict(session_key,
**insertion_loc_key,
insertion_number=1,
probe=probe,
electrode_config_name=electrode_config_name),
ignore_extra_fields=True)
ephys.ProbeInsertion.ElectrodeSitePosition.insert(
(dict(session_key,
insertion_number=1,
probe=probe,
electrode_config_name=electrode_config_name,
electrode_group=0,
electrode=site_idx + 1,
electrode_posx=x * 1000,
electrode_posy=y * 1000,
electrode_posz=z * 1000)
for site_idx, (
x, y,
z) in enumerate(meta_data.extracellular.siteLocations)),
ignore_extra_fields=True)
print(
f'\tInsert ProbeInsertion - Location: {brain_location_key["brain_location_name"]}'
)
# ==================== Virus ====================
if 'virus' in meta_data._fieldnames and isinstance(
meta_data.virus, sio.matlab.mio5_params.mat_struct):
virus_info = dict(
virus_source=meta_data.virus.virusSource,
virus=meta_data.virus.virusID,
virus_lot_number=meta_data.virus.virusLotNumber
if len(meta_data.virus.virusLotNumber) != 0 else '',
virus_titer=meta_data.virus.virusTiter.replace('x10', '')
if meta_data.virus.virusTiter != 'untitered' else None)
virus.Virus.insert1(virus_info, skip_duplicates=True)
# -- BrainLocation
brain_location_key = (experiment.BrainLocation & {
'brain_area': meta_data.virus.infectionLocation,
'hemisphere': hemi,
'skull_reference': skull_reference
}).fetch1('KEY')
virus_injection = dict(
{
**virus_info,
**subject_key,
**brain_location_key
},
injection_date=parse_date(meta_data.virus.injectionDate))
virus.VirusInjection.insert([
dict(virus_injection,
injection_id=inj_idx + 1,
ml_location=coord[0] * 1000,
ap_location=coord[1] * 1000,
dv_location=coord[2] * 1000,
injection_volume=vol)
for inj_idx, (coord, vol) in enumerate(
zip(meta_data.virus.infectionCoordinates,
meta_data.virus.injectionVolume))
],
ignore_extra_fields=True,
skip_duplicates=True)
print(
f'\tInsert Virus Injections - Count: {len(meta_data.virus.injectionVolume)}'
)
# ==================== Photostim ====================
if 'photostim' in meta_data._fieldnames and isinstance(
meta_data.photostim, sio.matlab.mio5_params.mat_struct):
photostimLocation = (
meta_data.photostim.photostimLocation if isinstance(
meta_data.photostim.photostimLocation, np.ndarray) else
np.array([meta_data.photostim.photostimLocation]))
photostimCoordinates = (
meta_data.photostim.photostimCoordinates if isinstance(
meta_data.photostim.photostimCoordinates[0], np.ndarray)
else np.array([meta_data.photostim.photostimCoordinates]))
photostim_locs = []
for ba in set(photostimLocation):
coords = photostimCoordinates[photostimLocation == ba]
for coord in coords:
photostim_locs.append(
(ba, 'left' if coord[1] < 0 else 'right', coord))
if len(coords) > 1:
photostim_locs.append(
(ba, 'both',
np.array(
[coords[0][0],
abs(coords[0][1]), coords[0][2]])))
for stim_idx, (loc, hem, coord) in enumerate(photostim_locs):
brain_location_key = (experiment.BrainLocation & dict(
brain_area=loc,
hemisphere=hem,
skull_reference=skull_reference)).fetch1('KEY')
experiment.Photostim.insert1(dict(
session_key,
**brain_location_key,
photo_stim=stim_idx + 1,
photostim_device=photostim_devices[
meta_data.photostim.photostimWavelength],
ml_location=coord[0] * 1000,
ap_location=coord[1] * 1000,
dv_location=coord[2] * 1000),
ignore_extra_fields=True)
print(
f'\tInsert Photostim - Count: {len(meta_data.photostim.photostimLocation)}'
)
def main(data_dir='/data/data_structure'):
data_dir = pathlib.Path(data_dir)
if not data_dir.exists():
raise FileNotFoundError(f'Path not found!! {data_dir.as_posix()}')
# ==================== DEFINE CONSTANTS =====================
trial_type_str = ['HitL', 'HitR', 'ErrL', 'ErrR', 'NoLickL', 'NoLickR']
trial_type_mapper = {
'HitR': ('hit', 'right'),
'HitL': ('hit', 'left'),
'ErrR': ('miss', 'right'),
'ErrL': ('miss', 'left'),
'NoLickR': ('ignore', 'right'),
'NoLickL': ('ignore', 'left')
}
cell_type_mapper = {
'p': 'PT',
'i': 'IT',
'': 'N/A',
'in': 'interneuron',
'pn': 'Pyr'
}
post_resp_tlim = 2 # a trial may last at most 2 seconds after response cue
task_protocol = {'task': 'audio delay', 'task_protocol': 1}
insert_kwargs = {
'ignore_extra_fields': True,
'allow_direct_insert': True,
'skip_duplicates': True
}
# ================== INGESTION OF DATA ==================
data_files = data_dir.glob('*.mat')
for data_file in data_files:
print(f'-- Read {data_file} --')
fname = data_file.stem
subject_nickname = re.search('data_(an\d+)_', fname).group(1)
session_date = parse_date(
re.search(subject_nickname + '_(\d+_\d+_\d+_)',
fname).group(1).replace('_', ''))
depth = int(re.search('_(\d+)(_fv|$)', fname).group(1))
fov = int(re.search('fv(\d+)', fname).group(1)) if re.search(
'fv(\d+)', fname) else 1
sessions = (experiment.Session & (lab.Subject & {
'subject_nickname': subject_nickname
}).proj() & {
'session_date': session_date,
'fov': fov
} & (experiment.Session.ImagingDepth & {
'imaging_depth': depth
}))
if len(sessions) < 2:
session_key = sessions.fetch1('KEY')
else:
raise Exception('Multiple sessions found for {fname}')
print(f'\tMatched: {session_key}')
if imaging.TrialTrace & session_key:
print('Data ingested, skipping over...')
continue
sess_data = sio.loadmat(data_file,
struct_as_record=False,
squeeze_me=True)['obj']
# ---- trial data ----
trial_zip = zip(sess_data.trialIds, sess_data.trialStartTimes,
sess_data.trialTypeMat[:6, :].T,
sess_data.trialPropertiesHash.value[0],
sess_data.trialPropertiesHash.value[1],
sess_data.trialPropertiesHash.value[2])
print('---- Ingesting trial data ----')
(session_trials, behavior_trials, trial_events) = [], [], []
for (tr_id, tr_start, trial_type_mtx, sample_start, delay_start,
response_start) in tqdm(trial_zip):
tkey = dict(session_key,
trial=tr_id,
start_time=Decimal(tr_start),
stop_time=Decimal(tr_start + response_start +
post_resp_tlim))
session_trials.append(tkey)
trial_type = np.array(trial_type_str)[trial_type_mtx.astype(bool)]
if len(trial_type) == 1:
outcome, trial_instruction = trial_type_mapper[trial_type[0]]
else:
outcome, trial_instruction = 'non-performing', 'non-performing'
bkey = dict(tkey,
**task_protocol,
trial_instruction=trial_instruction,
outcome=outcome,
early_lick='no early')
behavior_trials.append(bkey)
for etype, etime in zip(
('sample', 'delay', 'go'),
(sample_start, delay_start, response_start)):
if not np.isnan(etime):
trial_events.append(
dict(tkey,
trial_event_id=len(trial_events) + 1,
trial_event_type=etype,
trial_event_time=etime))
# insert trial info
experiment.SessionTrial.insert(session_trials, **insert_kwargs)
experiment.BehaviorTrial.insert(behavior_trials, **insert_kwargs)
experiment.TrialEvent.insert(trial_events, **insert_kwargs)
# ---- Scan info ----
scan = dict(**session_key,
image_gcamp=sess_data.images.value[0],
image_ctb=sess_data.images.value[1],
image_beads=sess_data.images.value[2]
if len(sess_data.images.value[2]) else np.nan,
frame_time=sess_data.timeSeriesArrayHash.value[0].time)
cell_type_vec = [
ct if len(ct) else ''
for ct in sess_data.timeSeriesArrayHash.value[0].cellType
]
rois = [
dict(**scan,
roi_idx=idx,
cell_type=cell_type_mapper[cell_type],
roi_trace=roi_trace,
neuropil_trace=neuropil_trace,
roi_pixel_list=roi_plist,
neuropil_pixel_list=neuropil_plist,
inc=bool(np.mean(roi_trace) / np.mean(neuropil_trace) > 1.05))
for (idx, cell_type, roi_trace, neuropil_trace, roi_plist,
neuropil_plist) in zip(
sess_data.timeSeriesArrayHash.value[0].ids, cell_type_vec,
sess_data.timeSeriesArrayHash.value[0].valueMatrix,
sess_data.timeSeriesArrayHash.value[1].valueMatrix,
sess_data.timeSeriesArrayHash.value[0].pixel_list,
sess_data.timeSeriesArrayHash.value[1].pixel_list)
]
imaging.Scan.insert1(scan, **insert_kwargs)
imaging.Scan.Roi.insert(rois, **insert_kwargs)
tr_events = {
tr: (float(stime), float(gotime))
for tr, stime, gotime in zip(
*(experiment.SessionTrial * experiment.TrialEvent
& session_key & 'trial_event_type = "go"'
).fetch('trial', 'start_time', 'trial_event_time'))
}
print('---- Ingesting trial trace ----')
trials = sess_data.timeSeriesArrayHash.value[0].trial
frame_time = sess_data.timeSeriesArrayHash.value[0].time
trial_traces = []
for roi_id, trace in tqdm(
zip(sess_data.timeSeriesArrayHash.value[0].ids,
sess_data.timeSeriesArrayHash.value[0].valueMatrix)):
for tr in set(trials):
if tr in tr_events:
go_cue_time = sum(tr_events[tr])
go_id = np.abs(frame_time - go_cue_time).argmin()
trial_traces += [
dict(**scan,
roi_idx=roi_id,
trial=tr,
original_time=frame_time[go_id - 45:go_id + 45],
aligned_time=frame_time[go_id - 45:go_id + 45] -
go_cue_time,
aligned_trace=trace[go_id - 45:go_id + 45],
dff=(trace[go_id - 45:go_id + 45] -
np.mean(trace[go_id - 45:go_id - 39])) /
np.mean(trace[go_id - 45:go_id - 39]))
]
imaging.TrialTrace.insert(trial_traces, **insert_kwargs)
def main(data_dir='./data/data_structure'):
data_dir = pathlib.Path(data_dir)
if not data_dir.exists():
raise FileNotFoundError(f'Path not found!! {data_dir.as_posix()}')
# ==================== DEFINE CONSTANTS =====================
session_suffixes = ['a', 'b', 'c', 'd', 'e']
trial_type_str = ['HitR', 'HitL', 'ErrR', 'ErrL', 'NoLickR', 'NoLickL']
trial_type_mapper = {
'HitR': ('hit', 'right'),
'HitL': ('hit', 'left'),
'ErrR': ('miss', 'right'),
'ErrL': ('miss', 'left'),
'NoLickR': ('ignore', 'right'),
'NoLickL': ('ignore', 'left')
}
photostim_mapper = {
1: {
'brain_area': 'alm',
'hemi': 'left',
'duration': 0.5,
'spot': 1,
'pre_go_end_time': 1.6,
'period': 'sample'
},
2: {
'brain_area': 'alm',
'hemi': 'left',
'duration': 0.5,
'spot': 1,
'pre_go_end_time': 0.8,
'period': 'early_delay'
},
3: {
'brain_area': 'alm',
'hemi': 'left',
'duration': 0.5,
'spot': 1,
'pre_go_end_time': 0.3,
'period': 'middle_delay'
},
4: {
'brain_area': 'alm',
'hemi': 'left',
'duration': 0.8,
'spot': 1,
'pre_go_end_time': 0.9,
'period': 'early_delay'
},
5: {
'brain_area': 'alm',
'hemi': 'right',
'duration': 0.8,
'spot': 1,
'pre_go_end_time': 0.9,
'period': 'early_delay'
},
6: {
'brain_area': 'alm',
'hemi': 'both',
'duration': 0.8,
'spot': 4,
'pre_go_end_time': 0.9,
'period': 'early_delay'
},
7: {
'brain_area': 'alm',
'hemi': 'both',
'duration': 0.8,
'spot': 1,
'pre_go_end_time': 0.9,
'period': 'early_delay'
},
8: {
'brain_area': 'alm',
'hemi': 'left',
'duration': 0.8,
'spot': 4,
'pre_go_end_time': 0.9,
'period': 'early_delay'
},
9: {
'brain_area': 'alm',
'hemi': 'right',
'duration': 0.8,
'spot': 4,
'pre_go_end_time': 0.9,
'period': 'early_delay'
}
}
cell_type_mapper = {'pyramidal': 'Pyr', 'FS': 'FS', 'IT': 'IT', 'PT': 'PT'}
post_resp_tlim = 2 # a trial may last at most 2 seconds after response cue
task_protocol = {'task': 'audio delay', 'task_protocol': 1}
clustering_method = 'manual'
insert_kwargs = {
'ignore_extra_fields': True,
'allow_direct_insert': True,
'skip_duplicates': True
}
# ================== INGESTION OF DATA ==================
data_files = data_dir.glob('*.mat')
for data_file in data_files:
print(f'-- Read {data_file} --')
fname = data_file.stem
subject_id = int(re.search('ANM\d+', fname).group().replace('ANM', ''))
session_date = parse_date(
re.search('_\d+', fname).group().replace('_', ''))
sessions = (experiment.Session & {
'subject_id': subject_id,
'session_date': session_date
})
if len(sessions) < 2:
session_key = sessions.fetch1('KEY')
else:
if fname[-1] in session_suffixes:
sess_num = sessions.fetch('session', order_by='session')
session_letter_mapper = {
letter: s_no
for letter, s_no in zip(session_suffixes, sess_num)
}
session_key = (sessions & {
'session': session_letter_mapper[fname[-1]]
}).fetch1('KEY')
else:
raise Exception('Multiple sessions found for {fname}')
print(f'\tMatched: {session_key}')
if ephys.TrialSpikes & session_key:
print('Data ingested, skipping over...')
continue
sess_data = sio.loadmat(data_file,
struct_as_record=False,
squeeze_me=True)['obj']
# get time conversion factor - (-1) to take into account Matlab's 1-based indexing
ts_time_conversion = time_unit_conversion_factor[
sess_data.timeUnitNames[
sess_data.timeSeriesArrayHash.value.timeUnit - 1]]
trial_time_conversion = time_unit_conversion_factor[
sess_data.timeUnitNames[sess_data.trialTimeUnit - 1]]
unit_time_converstion = time_unit_conversion_factor[
sess_data.timeUnitNames[sess_data.eventSeriesHash.value[0].timeUnit
- 1]]
# ---- time-series data ----
ts_tvec = sess_data.timeSeriesArrayHash.value.time * ts_time_conversion
ts_trial = sess_data.timeSeriesArrayHash.value.trial
lick_trace = sess_data.timeSeriesArrayHash.value.valueMatrix[:, 0]
aom_input_trace = sess_data.timeSeriesArrayHash.value.valueMatrix[:, 1]
laser_power = sess_data.timeSeriesArrayHash.value.valueMatrix[:, 2]
# ---- trial data ----
photostims = (experiment.Photostim * experiment.BrainLocation
& session_key)
trial_zip = zip(
sess_data.trialIds,
sess_data.trialStartTimes * trial_time_conversion,
sess_data.trialTypeMat[:6, :].T, sess_data.trialTypeMat[6, :].T,
sess_data.trialPropertiesHash.value[0] * trial_time_conversion,
sess_data.trialPropertiesHash.value[1] * trial_time_conversion,
sess_data.trialPropertiesHash.value[2] * trial_time_conversion,
sess_data.trialPropertiesHash.value[-1])
print('---- Ingesting trial data ----')
(session_trials, behavior_trials, trial_events, photostim_trials,
photostim_events, photostim_traces,
lick_traces) = [], [], [], [], [], [], []
for (tr_id, tr_start, trial_type_mtx, is_early_lick, sample_start,
delay_start, response_start, photostim_type) in tqdm(trial_zip):
tkey = dict(
session_key,
trial=tr_id,
start_time=Decimal(tr_start),
stop_time=Decimal(tr_start + (
0 if np.isnan(response_start) else response_start) +
post_resp_tlim))
session_trials.append(tkey)
trial_type = np.array(trial_type_str)[trial_type_mtx.astype(bool)]
if len(trial_type) == 1:
outcome, trial_instruction = trial_type_mapper[trial_type[0]]
else:
outcome, trial_instruction = 'non-performing', 'non-performing'
bkey = dict(tkey,
**task_protocol,
trial_instruction=trial_instruction,
outcome=outcome,
early_lick='early' if is_early_lick else 'no early')
behavior_trials.append(bkey)
lick_traces.append(
dict(bkey,
lick_trace=lick_trace[ts_trial == tr_id],
lick_trace_timestamps=ts_tvec[ts_trial == tr_id] -
tr_start))
for etype, etime in zip(
('sample', 'delay', 'go'),
(sample_start, delay_start, response_start)):
if not np.isnan(etime):
trial_events.append(
dict(tkey,
trial_event_id=len(trial_events) + 1,
trial_event_type=etype,
trial_event_time=etime))
if photostims and photostim_type != 0:
pkey = dict(tkey)
photostim_trials.append(pkey)
photostim_type = photostim_type.astype(int)
if photostim_type in photostim_mapper:
photstim_detail = photostim_mapper[photostim_type]
photostim_key = (
photostims & {
'brain_area': photstim_detail['brain_area'],
'hemisphere': photstim_detail['hemi']
})
if photostim_key:
photostim_key = photostim_key.fetch1('KEY')
stim_power = laser_power[ts_trial == tr_id]
stim_power = np.where(
np.isinf(stim_power), 0,
stim_power) # handle cases where stim power is Inf
photostim_events.append(
dict(pkey,
**photostim_key,
photostim_event_id=len(photostim_events) + 1,
power=stim_power.max()
if len(stim_power) > 0 else None,
duration=Decimal(photstim_detail['duration']),
photostim_event_time=response_start -
photstim_detail['pre_go_end_time'] -
photstim_detail['duration'],
stim_spot_count=photstim_detail['spot'],
photostim_period=photstim_detail['period']))
photostim_traces.append(
dict(
pkey,
aom_input_trace=aom_input_trace[ts_trial ==
tr_id],
laser_power=laser_power[ts_trial == tr_id],
photostim_timestamps=ts_tvec[ts_trial == tr_id]
- tr_start))
# insert trial info
experiment.SessionTrial.insert(session_trials, **insert_kwargs)
experiment.BehaviorTrial.insert(behavior_trials, **insert_kwargs)
experiment.PhotostimTrial.insert(photostim_trials, **insert_kwargs)
experiment.TrialEvent.insert(trial_events, **insert_kwargs)
experiment.PhotostimEvent.insert(photostim_events, **insert_kwargs)
experiment.PhotostimTrace.insert(photostim_traces, **insert_kwargs)
tracking.LickTrace.insert(lick_traces, **insert_kwargs)
# ---- units ----
insert_key = (ephys.ProbeInsertion & session_key).fetch1()
ap, dv = (ephys.ProbeInsertion.InsertionLocation & session_key).fetch1(
'ap_location', 'dv_location')
e_sites = {
e: (y - ap, z - dv)
for e, y, z in zip(
*(ephys.ProbeInsertion.ElectrodeSitePosition & session_key
).fetch('electrode', 'electrode_posy', 'electrode_posz'))
}
tr_events = {
tr: (float(stime), float(gotime))
for tr, stime, gotime in zip(
*(experiment.SessionTrial * experiment.TrialEvent
& session_key & 'trial_event_type = "go"'
).fetch('trial', 'start_time', 'trial_event_time'))
}
print('---- Ingesting spike data ----')
unit_spikes, unit_cell_types, trial_spikes = [], [], []
for u_name, u_value in tqdm(
zip(sess_data.eventSeriesHash.keyNames,
sess_data.eventSeriesHash.value)):
unit = int(re.search('\d+', u_name).group())
electrode = np.unique(u_value.channel)[0]
spike_times = u_value.eventTimes * unit_time_converstion
unit_key = dict(insert_key,
clustering_method=clustering_method,
unit=unit)
unit_spikes.append(
dict(unit_key,
electrode_group=0,
unit_quality='good',
electrode=electrode,
unit_posx=e_sites[electrode][0],
unit_posy=e_sites[electrode][1],
spike_times=spike_times,
waveform=u_value.waveforms))
unit_cell_types += [
dict(unit_key,
cell_type=(cell_type_mapper[cell_type]
if len(cell_type) > 0 else 'N/A'))
for cell_type in (
u_value.cellType if isinstance(u_value.cellType, (
list, np.ndarray)) else [u_value.cellType])
]
# get trial's spike times, shift by start-time, then by go-time -> align to go-time
trial_spikes += [
dict(unit_key,
trial=tr,
spike_times=(spike_times[u_value.eventTrials == tr] -
tr_events[tr][0] - tr_events[tr][1]))
for tr in set(u_value.eventTrials) if tr in tr_events
]
ephys.Unit.insert(unit_spikes, **insert_kwargs)
ephys.UnitCellType.insert(unit_cell_types, **insert_kwargs)
ephys.TrialSpikes.insert(trial_spikes, **insert_kwargs)
def main(meta_data_dir='/data/meta_data'):
meta_data_dir = pathlib.Path(meta_data_dir)
if not meta_data_dir.exists():
raise FileNotFoundError(f'Path not found!! {meta_data_dir.as_posix()}')
# ==================== DEFINE CONSTANTS =====================
# ---- inferred from paper ----
hemi = 'left'
skull_reference = 'bregma'
photostim_devices = {473: 'LaserGem473', 594: 'LaserCoboltMambo100', 596: 'LaserCoboltMambo100'}
# ---- from lookup ----
probe = 'A4x8-5mm-100-200-177'
electrode_config_name = 'silicon32'
# ---- virus source mapper ----
virus_source_mapper = {
'Upenn core': 'UPenn',
'Upenn Core': 'UPenn',
'UPenn': 'UPenn'
}
# ---- brain location mapper ----
brain_location_mapper = {
'Left ALM': 'left_alm',
'Left pontine nucleus': 'left_pons'
}
# ================== INGESTION OF METADATA ==================
# ---- delete all Sessions ----
experiment.Session.delete()
# ---- insert metadata ----
meta_data_files = meta_data_dir.glob('*.mat')
for meta_data_file in tqdm(meta_data_files):
print(f'-- Read {meta_data_file} --')
meta_data = sio.loadmat(meta_data_file, struct_as_record = False, squeeze_me=True)['meta_data']
# ==================== person ====================
person_key = dict(username=meta_data.experimenters,
fullname=meta_data.experimenters)
lab.Person.insert1(person_key, skip_duplicates=True)
# ==================== subject gene modification ====================
modified_genes = (meta_data.animalGeneModification
if isinstance(meta_data.animalGeneModification, (np.ndarray, list))
else [meta_data.animalGeneModification])
lab.ModifiedGene.insert((dict(gene_modification=g, gene_modification_description=g)
for g in modified_genes), skip_duplicates=True)
# ==================== subject strain ====================
animal_strains = (meta_data.animalStrain
if isinstance(meta_data.animalStrain, (np.ndarray, list))
else [meta_data.animalStrain])
lab.AnimalStrain.insert(zip(animal_strains), skip_duplicates=True)
# ==================== subject ====================
animal_id = (meta_data.animalID[0]
if isinstance(meta_data.animalID, (np.ndarray, list)) else meta_data.animalID)
animal_source = (meta_data.animalSource[0]
if isinstance(meta_data.animalSource, (np.ndarray, list)) else meta_data.animalSource)
subject_key = dict(subject_id=int(re.search('\d+', animal_id).group()),
subject_nickname=re.search('meta_(an\d+)_', meta_data_file.stem).group(1),
sex=meta_data.sex[0].upper() if len(meta_data.sex) != 0 else 'U',
species=meta_data.species,
animal_source=animal_source)
try:
date_of_birth = parse_date(meta_data.dateOfBirth)
subject_key['date_of_birth'] = date_of_birth
except:
pass
lab.AnimalSource.insert1((animal_source,), skip_duplicates=True)
with lab.Subject.connection.transaction:
if subject_key not in lab.Subject.proj():
lab.Subject.insert1(subject_key)
lab.Subject.GeneModification.insert((dict(subject_key, gene_modification=g) for g in modified_genes),
ignore_extra_fields=True)
lab.Subject.Strain.insert((dict(subject_key, animal_strain=strain) for strain in animal_strains),
ignore_extra_fields=True)
# ==================== session ====================
session_key = dict(subject_key, username=person_key['username'],
session=len(experiment.Session & subject_key) + 1,
session_date=parse_date(meta_data.dateOfExperiment))
filename = meta_data_file.stem
if re.search('fv(\d+)', filename):
session_key.update(fov=int(re.search('fv(\d+)', filename).group(1)))
experiment.Session.insert1(session_key, ignore_extra_fields=True)
experiment.Session.ImagingDepth.insert1(
dict(session_key, imaging_depth=int(re.search('_(\d+)(_fv|$)', filename).group(1))),
ignore_extra_fields=True)
print(f'\tInsert Session - {session_key["subject_id"]} - {session_key["session_date"]}')
# ==================== Virus ====================
if 'virus' in meta_data._fieldnames and isinstance(meta_data.virus, sio.matlab.mio5_params.mat_struct):
virus_info = dict(
virus_source=virus_source_mapper[meta_data.virus.Source],
virus=meta_data.virus.ID[1],
virus_titer=meta_data.virus.Titer.replace('x10', '') if meta_data.virus.Titer != 'unknown' else None)
virus.Virus.insert1(virus_info, skip_duplicates=True)
# -- BrainLocation
brain_location_key = (experiment.BrainLocation & {'brain_location_name': brain_location_mapper[meta_data.virus.Location],
'hemisphere': hemi,
'skull_reference': skull_reference}).fetch1('KEY')
virus_injection = dict(
{**virus_info, **subject_key, **brain_location_key},
injection_date=parse_date(meta_data.virus.injectionDate))
virus.VirusInjection.insert([dict(virus_injection,
injection_id=inj_idx + 1,
virus_dilution=float(re.search('1:(\d+) dilution',
meta_data.virus.Concentration).group(1)) \
if 'Concentraion' in meta_data.virus._fieldnames else None,
ml_location=coord[0] * 1000,
ap_location=coord[1] * 1000,
dv_location=coord[2] * 1000,
injection_volume=vol)
for inj_idx, (coord, vol) in enumerate(zip(meta_data.virus.Coordinates,
meta_data.virus.injectionVolume))],
ignore_extra_fields=True, skip_duplicates=True)
print(f'\tInsert Virus Injections - Count: {len(meta_data.virus.injectionVolume)}')
