def plot_consistency_bytrialtype(h5fname,
dfSummary,
dropChannels=None,
performance=None,
datatype=None,
trialTypes=None,
kind='point',
fisherThr=0.1,
limits=None):
pidTypes = ['unique', 'syn', 'red']
limitKWargs = {
'vmin': limits[0],
'vmax': limits[1]
} if limits is not None else {}
if performance is None:
dfSummaryEff = pd_query(dfSummary, {'datatype': datatype})
else:
dfSummaryEff = pd_query(dfSummary, {
'datatype': datatype,
'performance': performance
})
dfColumns = ['mousename', 'phase', 'consistency']
dfConsistencyDict = {
pidType: pd.DataFrame(columns=dfColumns)
for pidType in pidTypes
}
for (mousename, phase), df1 in dfSummaryEff.groupby(['mousename',
'phase']):
fnameSuffix = '_'.join([mousename, datatype, phase, str(performance)])
trialTypes = trialTypes if trialTypes is not None else sorted(
list(set(df1['trialType'])))
nTrialTypes = len(trialTypes)
dfTrialTypeDict = {}
for iPid, pidType in enumerate(pidTypes):
dfTrialTypeDict[pidType] = pd.DataFrame()
for trialType, dfTrialType in df1.groupby(['trialType']):
if trialType in trialTypes:
dfTmp = read_parse_joint_dataframe(dfTrialType,
h5fname, [mousename],
pidTypes,
dropChannels=dropChannels)
for iPid, pidType in enumerate(pidTypes):
dfTrialTypeDict[pidType][trialType] = dfTmp[pidType][
'muTrue_' + mousename]
for iPid, pidType in enumerate(pidTypes):
maxRange = 0.35 if pidType == 'syn' else 1.0
# As consistency metric perform Fischer's exact test for significant vs unsignificant links
# As pairplot show contingency tables
if kind == 'fisher':
fischerLabels = ['low', 'high']
rezMat = np.full((nTrialTypes, nTrialTypes), np.nan)
fig, ax = plt.subplots(nrows=nTrialTypes,
ncols=nTrialTypes,
figsize=(4 * nTrialTypes,
4 * nTrialTypes))
for idxTTi, iTT in enumerate(trialTypes):
ax[idxTTi, 0].set_ylabel(iTT)
ax[-1, idxTTi].set_xlabel(iTT)
for idxTTj, jTT in enumerate(trialTypes):
if idxTTi == idxTTj:
ax[idxTTi][idxTTj].hist(
dfTrialTypeDict[pidType][iTT],
range=[0, maxRange],
bins=50)
ax[idxTTi][idxTTj].axvline(x=fisherThr,
linestyle='--',
color='pink')
else:
iBin = dfTrialTypeDict[pidType][iTT] >= fisherThr
jBin = dfTrialTypeDict[pidType][jTT] >= fisherThr
M = confusion_matrix(iBin, jBin)
M = M.astype(float) / np.sum(M)
# consistency = fisher_exact(M, alternative='two_sided')[0]
# consistency = -np.log10(fisher_exact(M, alternative='two_sided')[1])
consistency = cohen_kappa_score(iBin, jBin)
rezMat[idxTTi][idxTTj] = consistency
sns.heatmap(ax=ax[idxTTi][idxTTj],
data=M,
annot=True,
cmap='jet',
xticklabels=fischerLabels,
yticklabels=fischerLabels)
else:
# As consistency metric use correlation coefficient between values
rezMat = np.zeros((nTrialTypes, nTrialTypes))
for idxTTi, iTT in enumerate(trialTypes):
for idxTTj, jTT in enumerate(trialTypes):
rezMat[idxTTi][idxTTj] = np.corrcoef(
dfTrialTypeDict[pidType][iTT],
dfTrialTypeDict[pidType][jTT])[0, 1]
if kind == 'point':
# As pairplot use scatter
pPlot = sns.pairplot(data=dfTrialTypeDict[pidType],
vars=trialTypes) #, kind='kde')
elif kind == 'heatmap':
# As pairplot use heatmap of binned scatter points
fig, ax = plt.subplots(nrows=nTrialTypes,
ncols=nTrialTypes,
figsize=(4 * nTrialTypes,
4 * nTrialTypes))
for idxTTi, iTT in enumerate(trialTypes):
ax[idxTTi, 0].set_ylabel(iTT)
ax[-1, idxTTi].set_xlabel(iTT)
for idxTTj, jTT in enumerate(trialTypes):
if idxTTi == idxTTj:
ax[idxTTi][idxTTj].hist(
dfTrialTypeDict[pidType][iTT],
range=[0, maxRange],
bins=50)
else:
ax[idxTTi][idxTTj].hist2d(
dfTrialTypeDict[pidType][iTT],
dfTrialTypeDict[pidType][jTT],
range=[[0, maxRange], [0, maxRange]],
bins=[50, 50],
cmap='jet')
plt.savefig('pics/' + pidType + '_consistency_bymouse_scatter_' +
fnameSuffix + '.png')
plt.close()
fig, ax = plt.subplots()
sns.heatmap(ax=ax,
data=rezMat,
annot=True,
cmap='jet',
xticklabels=trialTypes,
yticklabels=trialTypes,
**limitKWargs)
# imshow(fig, ax, rezMat, haveColorBar=True, limits=[0,1], xTicks=mice, yTicks=mice, cmap='jet')
plt.savefig('pics/' + pidType + '_consistency_bymouse_metric_' +
fnameSuffix + '.png')
plt.close()
avgConsistency = np.round(np.mean(offdiag_1D(rezMat)), 2)
dfConsistencyDict[pidType] = pd_append_row(
dfConsistencyDict[pidType], [mousename, phase, avgConsistency])
for iPid, pidType in enumerate(pidTypes):
fig, ax = plt.subplots()
dfPivot = pd_pivot(dfConsistencyDict[pidType], *dfColumns)
sns.heatmap(data=dfPivot, ax=ax, annot=True, cmap='jet', **limitKWargs)
fig.savefig('pics/' + 'summary_' + pidType + '_consistency_metric_' +
datatype + '_' + str(performance) + '.png')
plt.close()
def plot_consistency_significant_activity_byphase(dataDB,
ds,
intervals,
minTrials=10,
performance=None,
dropChannels=None):
rows = ds.list_dsets_pd()
rows['mousename'] = [
dataDB.find_mouse_by_session(session) for session in rows['session']
]
dfColumns = ['datatype', 'trialType', 'consistency']
dfConsistency = pd.DataFrame(columns=dfColumns)
for (datatype,
trialType), rowsMouse in rows.groupby(['datatype', 'trialType']):
pSigDict = {}
for mousename, rowsSession in rowsMouse.groupby(['mousename']):
pSig = []
for session, rowsTrial in rowsSession.groupby(['session']):
if (performance is None) or dataDB.is_matching_performance(
session, performance, mousename=mousename):
assert intervals[0] in list(rowsTrial['intervName'])
assert intervals[1] in list(rowsTrial['intervName'])
dsetLabel1 = pd_is_one_row(
pd_query(rowsTrial,
{'intervName': intervals[0]}))[1]['dset']
dsetLabel2 = pd_is_one_row(
pd_query(rowsTrial,
{'intervName': intervals[1]}))[1]['dset']
data1 = ds.get_data(dsetLabel1)
data2 = ds.get_data(dsetLabel2)
nTrials1 = data1.shape[0]
nTrials2 = data2.shape[1]
if (nTrials1 < minTrials) or (nTrials2 < minTrials):
print(session, datatype, trialType, 'too few trials',
nTrials1, nTrials2, ';; skipping')
else:
nChannels = data1.shape[1]
if dropChannels is not None:
channelMask = np.ones(nChannels).astype(bool)
channelMask[dropChannels] = 0
data1 = data1[:, channelMask]
data2 = data2[:, channelMask]
nChannels = nChannels - len(dropChannels)
pvals = [
wilcoxon(data1[:, iCh],
data2[:, iCh],
alternative='two-sided')[1]
for iCh in range(nChannels)
]
# pSig += [(np.array(pvals) < 0.01).astype(int)]
pSig += [-np.log10(np.array(pvals))]
# pSigDict[mousename] = np.sum(pSig, axis=0)
pSigDict[mousename] = np.mean(pSig, axis=0)
mice = sorted(dataDB.mice)
nMice = len(mice)
corrCoef = np.zeros((nMice, nMice))
for iMouse, iName in enumerate(mice):
for jMouse, jName in enumerate(mice):
corrCoef[iMouse, jMouse] = np.corrcoef(pSigDict[iName],
pSigDict[jName])[0, 1]
sns.pairplot(data=pd.DataFrame(pSigDict), vars=mice)
prefixPath = 'pics/consistency/significant_activity/byphase/bymouse/'
make_path(prefixPath)
plt.savefig(prefixPath + datatype + '_' + trialType + '.svg')
plt.close()
fig2, ax2 = plt.subplots()
ax2.imshow(corrCoef, vmin=0, vmax=1)
imshow(fig2,
ax2,
corrCoef,
title='Significance Correlation',
haveColorBar=True,
limits=[0, 1],
xTicks=mice,
yTicks=mice)
prefixPath = 'pics/consistency/significant_activity/byphase/bymouse_corr/'
make_path(prefixPath)
plt.savefig(prefixPath + datatype + '_' + trialType + '.svg')
plt.close()
avgConsistency = np.round(np.mean(offdiag_1D(corrCoef)), 2)
dfConsistency = pd_append_row(dfConsistency,
[datatype, trialType, avgConsistency])
fig, ax = plt.subplots()
dfPivot = pd_pivot(dfConsistency, *dfColumns)
sns.heatmap(data=dfPivot, ax=ax, annot=True, vmax=1, cmap='jet')
prefixPath = 'pics/consistency/significant_activity/byphase/'
make_path(prefixPath)
fig.savefig(prefixPath + 'consistency_' + str(performance) + '.svg')
plt.close()
def plot_consistency_significant_activity_byaction(dataDB,
ds,
minTrials=10,
performance=None,
dropChannels=None,
metric='accuracy',
limits=None):
testFunc = test_metric_by_name(metric)
rows = ds.list_dsets_pd()
rows['mousename'] = [
dataDB.find_mouse_by_session(session) for session in rows['session']
]
dfColumns = ['datatype', 'phase', 'consistency']
dfConsistency = pd.DataFrame(columns=dfColumns)
for (datatype,
intervName), rowsMouse in rows.groupby(['datatype', 'intervName']):
pSigDict = {}
for mousename, rowsSession in rowsMouse.groupby(['mousename']):
pSig = []
for session, rowsTrial in rowsSession.groupby(['session']):
if (performance is None) or dataDB.is_matching_performance(
session, performance, mousename=mousename):
if len(rowsTrial) != 2:
print(mousename, session, rowsTrial)
raise ValueError('Expected exactly 2 rows')
dsetLabels = list(rowsTrial['dset'])
data1 = ds.get_data(dsetLabels[0])
data2 = ds.get_data(dsetLabels[1])
nTrials1 = data1.shape[0]
nTrials2 = data2.shape[1]
if (nTrials1 < minTrials) or (nTrials2 < minTrials):
print(session, datatype, intervName, 'too few trials',
nTrials1, nTrials2, ';; skipping')
else:
nChannels = data1.shape[1]
if dropChannels is not None:
channelMask = np.ones(nChannels).astype(bool)
channelMask[dropChannels] = 0
data1 = data1[:, channelMask]
data2 = data2[:, channelMask]
nChannels = nChannels - len(dropChannels)
pvals = [
testFunc(data1[:, iCh], data2[:, iCh])
for iCh in range(nChannels)
]
# pSig += [(np.array(pvals) < 0.01).astype(int)]
pSig += [-np.log10(np.array(pvals))]
# pSigDict[mousename] = np.sum(pSig, axis=0)
pSigDict[mousename] = np.mean(pSig, axis=0)
mice = sorted(pSigDict.keys())
nMice = len(mice)
corrCoef = np.zeros((nMice, nMice))
for iMouse, iName in enumerate(mice):
for jMouse, jName in enumerate(mice):
corrCoef[iMouse, jMouse] = np.corrcoef(pSigDict[iName],
pSigDict[jName])[0, 1]
plotSuffix = '_'.join([datatype, str(performance), intervName])
sns.pairplot(data=pd.DataFrame(pSigDict), vars=mice)
prefixPath = 'pics/consistency/significant_activity/byaction/bymouse/'
make_path(prefixPath)
plt.savefig(prefixPath + plotSuffix + '.svg')
plt.close()
fig2, ax2 = plt.subplots()
ax2.imshow(corrCoef, vmin=0, vmax=1)
imshow(fig2,
ax2,
corrCoef,
title='Significance Correlation',
haveColorBar=True,
limits=[0, 1],
xTicks=mice,
yTicks=mice)
prefixPath = 'pics/consistency/significant_activity/byaction/bymouse_corr/'
make_path(prefixPath)
plt.savefig(prefixPath + plotSuffix + '.svg')
plt.close()
avgConsistency = np.round(np.mean(offdiag_1D(corrCoef)), 2)
dfConsistency = pd_append_row(dfConsistency,
[datatype, intervName, avgConsistency])
fig, ax = plt.subplots()
dfPivot = pd_pivot(dfConsistency, *dfColumns)
sns.heatmap(data=dfPivot, ax=ax, annot=True, vmax=1, cmap='jet')
prefixPath = 'pics/consistency/significant_activity/byaction/'
make_path(prefixPath)
fig.savefig(prefixPath + 'consistency_' + str(performance) + '.svg')
plt.close()
def plot_pca_consistency(dataDB,
intervNames=None,
dropFirst=None,
dropChannels=None):
mice = sorted(dataDB.mice)
nMice = len(mice)
dfColumns = ['datatype', 'phase', 'consistency']
dfConsistency = pd.DataFrame(columns=dfColumns)
if intervNames is None:
intervNames = dataDB.get_interval_names()
for datatype in dataDB.get_data_types():
for intervName in intervNames:
fnameSuffix = datatype + '_' + intervName
print(fnameSuffix)
dataLst = []
for iMouse, mousename in enumerate(mice):
dataRSPLst = dataDB.get_neuro_data({'mousename': mousename},
datatype=datatype,
intervName=intervName)
print(set([d.shape[1:] for d in dataRSPLst]))
dataRSP = np.concatenate(dataRSPLst, axis=0)
dataRP = np.mean(dataRSP, axis=1)
dataRP = zscore(dataRP, axis=0)
if dropChannels is not None:
channelMask = np.ones(dataRP.shape[1]).astype(bool)
channelMask[dropChannels] = 0
dataRP = dataRP[:, channelMask]
dataLst += [dataRP]
fig, ax = plt.subplots(nrows=nMice,
ncols=nMice,
figsize=(4 * nMice, 4 * nMice))
rezMat = np.zeros((nMice, nMice))
for iMouse in range(nMice):
ax[iMouse][0].set_ylabel(mice[iMouse])
ax[-1][iMouse].set_xlabel(mice[iMouse])
for jMouse in range(nMice):
rezXY, rezYX, arrXX, arrXY, arrYY, arrYX = pca.paired_comparison(
dataLst[iMouse], dataLst[jMouse], dropFirst=dropFirst)
rezMat[iMouse][jMouse] = rezXY # np.mean([rezXY, rezYX])
ax[iMouse][jMouse].plot(arrXX, label='XX')
ax[iMouse][jMouse].plot(arrXY, label='XY')
ax[iMouse][jMouse].legend()
plt.savefig('pca_consistency_bymouse_evals_' + fnameSuffix +
'.svg')
plt.close()
fig, ax = plt.subplots()
imshow(fig,
ax,
rezMat,
haveColorBar=True,
limits=[0, 1],
xTicks=mice,
yTicks=mice)
plt.savefig('pca_consistency_bymouse_metric_' + fnameSuffix +
'.svg')
plt.close()
avgConsistency = np.round(np.mean(offdiag_1D(rezMat)), 2)
dfConsistency = pd_append_row(
dfConsistency, [datatype, intervName, avgConsistency])
fig, ax = plt.subplots()
dfPivot = pd_pivot(dfConsistency, *dfColumns)
sns.heatmap(data=dfPivot, ax=ax, annot=True, vmin=0, vmax=1, cmap='jet')
fig.savefig('consistency_coactivity_metric.svg')
plt.close()
def plot_corr_consistency_l1_phase(dataDB,
nDropPCA=None,
dropChannels=None,
performance=None,
datatype=None):
mice = sorted(dataDB.mice)
phases = dataDB.get_interval_names()
nPhases = len(phases)
dfColumns = ['mousename', 'trialtype', 'consistency']
dfConsistency = pd.DataFrame(columns=dfColumns)
for iMouse, mousename in enumerate(mice):
for trialType in dataDB.get_trial_type_names():
fnameSuffix = '_'.join(
[datatype, mousename, trialType,
str(performance)])
print(fnameSuffix)
corrLst = []
for intervName in phases:
kwargs = {
'datatype': datatype,
'intervName': intervName,
'trialType': trialType
}
if performance is not None:
kwargs['performance'] = performance
dataRSPLst = dataDB.get_neuro_data({'mousename': mousename},
zscoreDim='rs',
**kwargs)
dataRSP = np.concatenate(dataRSPLst, axis=0)
dataRP = np.mean(dataRSP, axis=1)
# dataRP = zscore(dataRP, axis=0)
if dropChannels is not None:
channelMask = np.ones(dataRP.shape[1]).astype(bool)
channelMask[dropChannels] = 0
dataRP = dataRP[:, channelMask]
if nDropPCA is not None:
dataRP = drop_PCA(dataRP, nDropPCA)
corrLst += [tril_1D(np.corrcoef(dataRP.T))]
# fig, ax = plt.subplots(nrows=nMice, ncols=nMice, figsize=(4 * nMice, 4 * nMice))
pairDict = {}
rezMat = np.zeros((nPhases, nPhases))
for idxNamei, iName in enumerate(phases):
pairDict[iName] = corrLst[idxNamei]
for idxNamej, jName in enumerate(phases):
rezMat[idxNamei][idxNamej] = np.corrcoef(
corrLst[idxNamei], corrLst[idxNamej])[0, 1]
pPlot = sns.pairplot(data=pd.DataFrame(pairDict),
vars=phases,
kind='kde')
prefixPath = 'pics/consistency/corr/phase/dropPCA_' + str(
nDropPCA) + '/scatter/'
make_path(prefixPath)
plt.savefig(prefixPath + 'scatter_' + fnameSuffix + '.svg')
plt.close()
fig, ax = plt.subplots()
imshow(fig,
ax,
rezMat,
haveColorBar=True,
limits=[0, 1],
xTicks=phases,
yTicks=phases,
cmap='jet')
prefixPath = 'pics/consistency/corr/phase/dropPCA_' + str(
nDropPCA) + '/metric/'
make_path(prefixPath)
plt.savefig(prefixPath + 'metric_' + fnameSuffix + '.svg')
plt.close()
avgConsistency = np.round(np.mean(offdiag_1D(rezMat)), 2)
dfConsistency = pd_append_row(
dfConsistency, [mousename, trialType, avgConsistency])
fig, ax = plt.subplots()
dfPivot = pd_pivot(dfConsistency, *dfColumns)
sns.heatmap(data=dfPivot, ax=ax, annot=True, vmin=0, vmax=1, cmap='jet')
prefixPath = 'pics/consistency/corr/phase/dropPCA_' + str(nDropPCA) + '/'
make_path(prefixPath)
fig.savefig(prefixPath + datatype + '_' + str(performance) + '.svg')
plt.close()
def plot_corr_consistency_l1_mouse(
dataDB,
nDropPCA=None,
dropChannels=None,
exclQueryLst=None,
**kwargs): # performances=None, trialTypes=None,
assert 'intervName' in kwargs.keys(), 'Requires phases'
dps = DataParameterSweep(dataDB,
exclQueryLst,
mousename='auto',
intervName='auto',
datatype='auto',
**kwargs)
mice = sorted(dataDB.mice)
nMice = len(mice)
for paramExtraVals, dfTmp in dps.sweepDF.groupby(
dps.invert_param(['mousename', 'datatype', 'intervName'])):
plotExtraSuffix = param_vals_to_suffix(paramExtraVals)
dfColumns = ['datatype', 'phase', 'consistency']
dfConsistency = pd.DataFrame(columns=dfColumns)
for paramVals, dfMouse in dfTmp.groupby(dps.invert_param(['mousename'
])):
plotSuffix = param_vals_to_suffix(paramVals)
print(plotSuffix)
corrLst = []
for idx, row in dfMouse.iterrows():
plotSuffix = '_'.join([str(s) for s in row.values])
print(plotSuffix)
kwargsThis = pd_row_to_kwargs(row,
parseNone=True,
dropKeys=['mousename'])
dataRSPLst = dataDB.get_neuro_data(
{
'mousename': row['mousename']
}, # NOTE: zscore channels for each session to avoid session-wise effects
zscoreDim='rs',
**kwargsThis)
dataRSP = np.concatenate(dataRSPLst, axis=0)
dataRP = np.mean(dataRSP, axis=1)
# dataRP = zscore(dataRP, axis=0)
if dropChannels is not None:
channelMask = np.ones(dataRP.shape[1]).astype(bool)
channelMask[dropChannels] = 0
dataRP = dataRP[:, channelMask]
if nDropPCA is not None:
dataRP = drop_PCA(dataRP, nDropPCA)
corrLst += [tril_1D(np.corrcoef(dataRP.T))]
# fig, ax = plt.subplots(nrows=nMice, ncols=nMice, figsize=(4 * nMice, 4 * nMice))
pairDict = {}
rezMat = np.zeros((nMice, nMice))
for iMouse, mousename in enumerate(mice):
# ax[iMouse][0].set_ylabel(mice[iMouse])
# ax[-1][iMouse].set_xlabel(mice[iMouse])
pairDict[mousename] = corrLst[iMouse]
for jMouse in range(nMice):
# rezMat[iMouse][jMouse] = 1 - rmae(corrLst[iMouse], corrLst[jMouse])
rezMat[iMouse][jMouse] = np.corrcoef(
corrLst[iMouse], corrLst[jMouse])[0, 1]
# cci = offdiag_1D(corrLst[iMouse])
# ccj = offdiag_1D(corrLst[jMouse])
# ax[iMouse][jMouse].plot(cci, ccj, '.')
pPlot = sns.pairplot(data=pd.DataFrame(pairDict),
vars=mice,
kind='kde')
prefixPath = 'pics/consistency/corr/mouse/dropPCA_' + str(
nDropPCA) + '/scatter/'
make_path(prefixPath)
plt.savefig(prefixPath + 'scatter_' + plotSuffix + '.svg')
plt.close()
fig, ax = plt.subplots()
imshow(fig,
ax,
rezMat,
haveColorBar=True,
limits=[0, 1],
xTicks=mice,
yTicks=mice,
cmap='jet')
prefixPath = 'pics/consistency/corr/mouse/dropPCA_' + str(
nDropPCA) + '/metric/'
make_path(prefixPath)
plt.savefig(prefixPath + 'metric_' + plotSuffix + '.svg')
plt.close()
avgConsistency = np.round(np.mean(offdiag_1D(rezMat)), 2)
dfConsistency = pd_append_row(
dfConsistency,
[row['datatype'], row['intervName'], avgConsistency])
fig, ax = plt.subplots()
dfPivot = pd_pivot(dfConsistency, *dfColumns)
sns.heatmap(data=dfPivot,
ax=ax,
annot=True,
vmin=0,
vmax=1,
cmap='jet')
prefixPath = 'pics/consistency/corr/mouse/dropPCA_' + str(
nDropPCA) + '/'
make_path(prefixPath)
fig.savefig(prefixPath + plotExtraSuffix + '.svg')
plt.close()