def run():
from oneibl.onelight import ONE, repository
# Get a session
one = ONE()
one.set_figshare_url = "https://figshare.com/articles/steinmetz/9974357"
sessions = one.search(['spikes'])
sess_n = 36
try:
sess = sessions[36]
except:
pass
# ----------------------------- Get metadata/data ---------------------------- #
df_fp = os.path.join('data', f'sess{sess_n}_alldata.h5')
if not os.path.isfile(df_fp):
# Download some data
channels_loc = one.load_dataset(sess, 'channels.brainLocation')
spikes_clust = one.load_dataset(sess, 'spikes.clusters')
spikes_times = (one.load_dataset(sess, 'spikes.times').ravel() *
1000).astype(np.int32) # convert to msecs
clust_probe = one.load_dataset(sess, 'clusters.probes')
clust_channel = one.load_dataset(sess, 'clusters.peakChannel')
print(
f"Regions in session: {sorted(set(channels_loc.allen_ontology))}")
# Assign a spike to each cluster and each cluster to a brain region
spikes = pd.DataFrame(
dict(cluster=spikes_clust.ravel(), times=spikes_times.ravel()))
spikes['channel'] = [
clust_channel[s[0]].ravel()[0] for s in spikes_clust
]
channels_loc_dic = {
i: row.allen_ontology
for i, row in channels_loc.iterrows()
}
spikes['region'] = [
channels_loc_dic[int(ch - 1)] for ch in spikes['channel']
]
print(spikes)
print(set(spikes.region.values))
spikes.to_hdf(df_fp, key='hdf')
else:
spikes = pd.read_hdf(df_fp)
# Get spikes for one brain region at the time
max_T = 10 * 60 * 1000 # 10 minutes
time = np.zeros((max_T, 1))
time_vals = np.arange(0, len(time))
sigma = 100 # 100ms std
for region in ['SCm']:
# prep some file paths
frates_fp = os.path.join('data', f'sess_{sess_n}_{region}_frates.npy')
reg = spikes.loc[spikes.region == region]
reg_clusters = list(set(reg.cluster.values))
print(f"Found {len(reg_clusters)} clusters for {region}")
# Create an array that is N_samples x N_cells and is 1 when a cell fires
print('Getting cells spikes')
cells = np.zeros((time.shape[0], len(reg_clusters)))
for n, clust in enumerate(reg_clusters):
clust_spikes = spikes.loc[(spikes.cluster == clust)
& (spikes.times < max_T)]
cells[clust_spikes.times, n] = 1
cells = cells[:max_T, :]
# Convert spikes to rates by summing with a gaussian kernel
print('Getting firing rates')
if not os.path.isfile(frates_fp):
frates = np.zeros_like(cells)
pool = mp.Pool(mp.cpu_count() - 2)
res = pool.map(get_cell_frate,
[(time, cells, cell, sigma)
for cell in np.arange(frates.shape[1])])
pool.close()
for n, rate in enumerate(res):
frates[:, n] = rate
np.save(frates_fp, frates)
else:
frates = np.load(frates_fp)
# TODO Bin spike counts in 1s bins for persistent homology
# embedd data into lower dimensional embeddiing spaces to facilitate future analysi
print('Reducing dimensionality')
pklfile = os.path.join('data', 'SCM_iso20.pkl')
if os.path.isfile(pklfile):
print(
'Skipping isomap because pickled isomap instance already exists'
)
else:
if N_samples < max_T:
idxs = choices(np.arange(max_T), k=N_samples)
sel_frates = frates[idxs, :]
else:
sel_frates = frates.copy()
stable_frates = np.sqrt(sel_frates)
iso_instance = manifold.Isomap(
5, 20, n_jobs=10
) # 5 nearest neightbours and 20 dimensional embedding
iso_instance = iso_instance.fit(stable_frates)
proj_data = iso_instance.transform(stable_frates)
np.save(os.path.join('data', 'SCm_iso_20.npy'), proj_data)
save_pickle(pklfile, iso_instance)
def __initialise_one_session():
one = ONE()
one.set_figshare_url(DataLoader.steinmetz_url)
return one
n_neurons = len(cluster_ids)
n_intervals = intervals.shape[0]
counts = np.zeros((n_neurons, n_intervals), dtype=np.uint32)
for j in range(n_intervals):
t0, t1 = intervals[j, :]
# Count the number of spikes in the window, for each neuron.
x = np.bincount(spike_clusters[intervals_idx[j, 0]:intervals_idx[j,
1]],
minlength=cluster_ids.max() + 1)
counts[:, j] = x[cluster_ids]
return counts, cluster_ids
# %%
# Set up ONE
one = ONE()
one.set_figshare_url('https://figshare.com/articles/steinmetz/9974357')
one.set_download_dir(join(DATA_PATH, '{subject}', '{date}'))
# Query all sessions with ephys and behavior
sessions = one.search(['spikes', 'trials'])
# Loop over recording sessions
timeconstant = pd.DataFrame()
for i, ses in enumerate(sessions):
print('Calculating time constants for session %d of %d' %
(i + 1, len(sessions)))
# Get subject and date
ses_path = normpath(ses)
ses_date = ses_path.split(os.sep)[-2]
scene.add_silhouette(spheres, lw=LW + 1)
# Interpolate cameras
anim.segment_fact = (end - framen) / (end - prev)
cam = anim._interpolate_cameras(cam1, cam2)
return cam
# ------------------------------- create scene ------------------------------- #
scene = Scene(inset=INSET, screenshots_folder="paper/screenshots")
scene.root._needs_silhouette = True
scene.add_brain_region("TH", "MOs", alpha=0.6, silhouette=True)
# download and process probe data
one = ONE()
one.set_figshare_url("https://figshare.com/articles/steinmetz/9974357")
# select session
sessions = one.search(["trials"])
sess = sessions[7] # 3
# Get spikes data
probes_locs = one.load_dataset(sess, "channels.brainLocation")
clu = one.load_dataset(sess, "clusters")
clu_probes = one.load_dataset(sess, "clusters.probes")
clu_channel = one.load_dataset(sess, "clusters.peakChannel")
spikes_clu = one.load_dataset(sess, "spikes.clusters")
spikes_times = one.load_dataset(sess, "spikes.times")
max_t = np.max(spikes_times)
# import modules # ============================================================================= from oneibl.onelight import ONE import numpy as np import matplotlib.pyplot as plt import seaborn from ibllib.misc import pprint from sklearn.linear_model import LogisticRegression from sklearn.model_selection import cross_val_score #%%============================================================================= # load data # ============================================================================= # create ONE object one = ONE() # list all dataset eids = one.search(['_ibl_trials.*']) # try one session eid = eids[10] # see all data set types dset_types = one.list(eid) # load a single dateset #f1 choice = one.load_dataset(eid, dset_types[0])
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
#%matplotlib inline
from oneibl.onelight import ONE
one = ONE()
one.set_figshare_url("https://figshare.com/articles/steinmetz/9974357")
sessions = one.search(['spikes'])
session = sessions[11]
one.list(session)
# loads all the objects from the session
trials = one.load_object(session, 'trials')
spikes = one.load_object(session, 'spikes')
clusters = one.load_object(session, 'clusters')
channels = one.load_object(session, 'channels')
probes = one.load_object(session, 'probes')
'''
See https://github.com/nsteinme/steinmetz-et-al-2019/wiki/data-files for description of dataset
Session 11 has:
channels: 1122
clusters: 1219
spikes: 11,757,530
trial timeline:
1. mouse holds wheel still for a short interval (0.2-0.5s)
2. trial initiates with stimulus onset
""" #%% ============================================================================= # import modules # ============================================================================= from oneibl.onelight import ONE import numpy as np import matplotlib.pyplot as plt import seaborn from ibllib.misc import pprint # ============================================================================= # load data # ============================================================================= # create ONE object one = ONE() # search terms one.search_terms() # list all dataset eids = one.search(['_ibl_trials.*']) # try one session eid = eids[9] # see all data set types dset_types = one.list(eid) # load a single dateset choice = one.load_dataset(eid, dset_types[0])
# %%
import pandas as pd
import numpy as np
from oneibl.onelight import ONE
one = ONE()
one.set_figshare_url = "https://figshare.com/articles/steinmetz/9974357"
sessions = one.search(['spikes'])
sess = sessions[2]
# print(one.list(sess))
# %%
channels_loc = one.load_dataset(sess, 'channels.brainLocation')
spikes_clust = one.load_dataset(sess, 'spikes.clusters')
spikes_times = one.load_dataset(sess, 'spikes.times')
clust_probe = one.load_dataset(sess, 'clusters.probes')
clust_channel = one.load_dataset(sess, 'clusters.peakChannel')
print(f"Regions in session: {set(sorted(channels_loc))}")
# %%
spikes = pd.DataFrame(dict(cluster=spikes_clust.ravel(), times=spikes_times.ravel()))
# %%
spikes['channel']= [clust_channel[s].ravel()[0] for s in spikes_clust]
spikes
# %%
cam = {
"pos": (-16170, -7127, 31776),
"viewup": (0, -1, 0),
"clippingRange": (27548, 67414),
"focalPoint": (7319, 2861, -3942),
"distance": 43901,
}
# create scene and edit root
scene = Scene(inset=INSET, screenshots_folder="paper/screenshots")
scene.root._needs_silhouette = True
scene.root._silhouette_kwargs["lw"] = 1
scene.root.alpha(0.2)
# download probe data from ONE
one = ONE()
one.set_figshare_url("https://figshare.com/articles/steinmetz/9974357")
# select sessions with trials
sessions = one.search(["trials"])
# get probe locations
probes_locs = []
for sess in sessions:
probes_locs.append(one.load_dataset(sess, "channels.brainLocation"))
# get single probe tracks
for locs in probes_locs:
k = int(len(locs) / 374.0)
for i in range(k):