forked from samsydco/HBN
-
Notifications
You must be signed in to change notification settings - Fork 0
/
event_ratings.py
234 lines (213 loc) · 9.99 KB
/
event_ratings.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
#!/usr/bin/env python3
import numpy as np
import pandas as pd
import deepdish as dd
import matplotlib.pyplot as plt
from settings import *
from event_comp import *
ev_conv = child_ev_conv
ev_figpath = figurepath+'event_annotations/'
# Tally of annotations at each time point
time = np.arange(0,nTR)
fig, (raw_ev_annot) = plt.subplots(figsize=(60, 20))
raw_ev_annot.hist(ev_annot, bins=nTR, linewidth=20,color='k')
yticks = list(np.arange(0, max(counts)+1, 2))
raw_ev_annot.spines['right'].set_color('none')
raw_ev_annot.spines['top'].set_color('none')
raw_ev_annot.set_xticks(np.append(time[0::nTR//5],time[-1]))
raw_ev_annot.set_xticklabels([str(int(s*TR//60))+':'+str(int(s*TR%60))+'0' for s in time][0::nTR//5]+['10:00'], fontsize=80)
raw_ev_annot.set_yticks(yticks)
plt.rcParams['xtick.labelsize']=50
plt.rcParams['ytick.labelsize']=80
raw_ev_annot.set_ylabel('# annotations', fontsize=80)
raw_ev_annot.set_xlabel('Time (minutes)', fontsize=80)
plt.tight_layout()
plt.savefig(ev_figpath+'ev_annots_hist.png', bbox_inches='tight')
# counts convolved with HRF
fig, (hrf_ann) = plt.subplots(figsize=(60, 20))
hrf_ann.plot(np.arange(nTR), ev_conv, linewidth=5,color='k')
hrf_ann.spines['right'].set_color('none')
hrf_ann.spines['top'].set_color('none')
plt.rcParams['xtick.labelsize']=80
plt.rcParams['ytick.labelsize']=80
hrf_ann.set_xlabel('TR (1 TR = 0.8 seconds)', fontsize=40)
plt.savefig(ev_figpath+'hrf_conv.png', bbox_inches='tight')
# illustration of counts with HRF-covlved counts
fig, (raw_ev_annot) = plt.subplots(figsize=(60, 20))
raw_ev_annot.hist(ev_annot, bins=nTR, linewidth=20,color='k')
raw_ev_annot.plot(np.arange(nTR), ev_conv, color='k',alpha=0.5,linewidth=5)
raw_ev_annot.spines['right'].set_color('none')
raw_ev_annot.spines['top'].set_color('none')
raw_ev_annot.set_xticks(np.append(time[0::nTR//5],time[-1]))
raw_ev_annot.set_xticklabels([str(int(s*TR//60))+':'+str(int(s*TR%60))+'0' for s in time][0::nTR//5]+['10:00'], fontsize=80)
raw_ev_annot.set_yticks(yticks)
plt.rcParams['xtick.labelsize']=50
plt.rcParams['ytick.labelsize']=80
raw_ev_annot.set_ylabel('# annotations', fontsize=80)
raw_ev_annot.set_xlabel('Time (minutes)', fontsize=80)
plt.tight_layout()
plt.savefig(ev_figpath+'ev_annots_both.png', bbox_inches='tight')
from HMM_settings import *
nTR = 750
D_ = dd.io.load(ISCpath+'HPC.h5',['/D'])[0]
D = np.zeros(nTR)
nsubj = 0
for b in range(nbinseq):
for bumps in D_[b].values():
D=D+bumps
nsubj+=1
D=D/nsubj
# average hippocampal activity overlaid with counts and ev_conv:
fig, (raw_ev_annot) = plt.subplots(figsize=(60, 20))
raw_ev_annot.hist(ev_annot, bins=nTR, linewidth=20,color='k')
raw_ev_annot.plot(np.arange(nTR), ev_conv, color='k',alpha=0.5,linewidth=5)
ax2 = raw_ev_annot.twinx()
ax2.plot(np.arange(nTR)[5:], D[5:], color='k',linewidth=10)
raw_ev_annot.spines['right'].set_color('none')
raw_ev_annot.spines['top'].set_color('none')
raw_ev_annot.set_xticks(np.append(time[0::nTR//5],time[-1]))
raw_ev_annot.set_xticklabels([str(int(s*TR//60))+':'+str(int(s*TR%60))+'0' for s in time][0::nTR//5]+['10:00'], fontsize=80)
raw_ev_annot.set_yticks(yticks)
plt.rcParams['xtick.labelsize']=50
plt.rcParams['ytick.labelsize']=80
raw_ev_annot.set_ylabel('# annotations', fontsize=80)
raw_ev_annot.set_xlabel('Time (minutes)', fontsize=80)
plt.tight_layout()
plt.savefig(ev_figpath+'ev_annots_HPC.png', bbox_inches='tight')
plt.savefig(ev_figpath+'ev_annots_HPC.eps', bbox_inches='tight')
# average hippocampal activity
fig, ax = plt.subplots(figsize=(60, 20))
ax.plot(np.arange(nTR)[5:], D[5:], color='k',linewidth=10)
ax.spines['right'].set_color('none')
ax.spines['top'].set_color('none')
ax.set_xticks(np.append(time[0::nTR//5],time[-1]))
ax.set_xticklabels([str(int(s*TR//60)) +':'+ str(int(s*TR%60))+'0' for s in time][0::nTR//5]+['10:00'], fontsize=80)
plt.rcParams['xtick.labelsize']=50
ax.set_xlabel('Time (minutes)', fontsize=80)
plt.tight_layout()
plt.savefig(figurepath+'HPC/HPC_timecourse.png', bbox_inches='tight')
# Compare annotations "ev_conv" with HPC bumps:
# Precidence for thinking this way comes from Aya Ben-Yakov's event saliency measure
sizedict = dd.io.load(ISCpath+'HPC_vol.h5')
sizedf = pd.DataFrame(data=sizedict)
import statsmodels.api as sm
import statsmodels.formula.api as sm2
import seaborn as sns
from scipy import stats
from statsmodels.stats.multitest import multipletests
from sklearn.model_selection import KFold
colors_age = ['#FCC3A1','#F08B63','#D02941','#70215D','#311638']
xticks = [str(int(round(eqbins[i])))+\
' - '+str(int(round(eqbins[i+1])))+' y.o.' for i in range(len(eqbins)-1)]
maxlag = 10
xcorrx = np.concatenate([np.arange(-maxlag,0)*TR,np.arange(maxlag+1)*TR])
nsub = nsubj
Dall = {}
bumplagdict = {'Age':[],'Time lag [s]':[],'Subj':[],'Exact_Age':[]}
for HPC in ['HPC','aHPC','pHPC']:
D = dd.io.load(ISCpath+HPC+'.h5','/D')
Dall[HPC] = D
bumplagdict['correlation_'+HPC] = []
for b in range(nbinseq):
for subj in Dall['HPC'][b].keys():
bumplagdict['Subj'].extend([subj]*len(xcorrx))
bumplagdict['Age'].extend([xticks[b]]*len(xcorrx))
bumplagdict['Exact_Age'].extend([Phenodf['Age'][Phenodf['EID'] == subj.split('/')[-1].split('.')[0].split('-')[1]].values[0]]*len(xcorrx))
for HPC in ['HPC','aHPC','pHPC']:
bumplagdict['correlation_'+HPC].extend(lag_pearsonr(ev_conv,Dall[HPC][b][subj],maxlag))
bumplagdict['Time lag [s]'].extend(xcorrx)
dfbumplag = pd.DataFrame(data=bumplagdict)
# Which time points post-0 are significantly different from zero?
times = dfbumplag['Time lag [s]'].unique()
tvals = np.zeros(len(times))
pvals = np.zeros(len(times))
for ti,tp in enumerate(times):
tvals[ti],pvals[ti] = stats.ttest_1samp(dfbumplag[dfbumplag['Time lag [s]']==tp]['correlation_HPC'],0)
pvals = pvals*len(pvals) # Bonferroni correction
best_t_i = times[np.argmin(pvals)]
best_t = 0 # Set from whole HPC
tempdf = dfbumplag[dfbumplag['Time lag [s]']==best_t]
t2 = np.zeros((3,len(xticks)))
p2 = np.zeros((3,len(xticks)))
for hi,HPC in enumerate(['HPC','aHPC','pHPC']):
for ai,age in enumerate(xticks):
t2[hi,ai],p2[hi,ai] = stats.ttest_1samp(tempdf[tempdf['Age']==age]['correlation_'+HPC],0)
tempsize = tempdf.merge(sizedf, on='Subj')
# test for linear vs polynomial fit of Age vs event_correlation:
# both line and U fit equally well!
nsplit = 20
y = [0]*int(np.floor(nsub/nsplit))*(nsplit-1)+[1]*int(np.floor(nsub/nsplit)+14)
kf = KFold(n_splits=nsplit, shuffle=True, random_state=2)
for HPC in ['HPC','aHPC','pHPC']:
sse = np.zeros((2,nsplit))
for degi,deg in enumerate([1,2]):
for split,Ls in enumerate(kf.split(np.arange(nsub),y)):
coeffs = np.polyfit(tempdf['Exact_Age'].iloc[Ls[0]],
tempdf['correlation_'+HPC].iloc[Ls[0]],deg)
pred = np.polyval(coeffs, tempdf['Exact_Age'].iloc[Ls[1]])
sse[degi,split] = np.sum(np.square(tempdf['correlation_'+HPC].iloc[Ls[1]]-pred))
print('Sum of Squared Error for Age-to-'+HPC+' to event boundary correlation is: '+
str(np.round(np.mean(sse[degi]),3))+'+/-'+str(np.round(np.std(sse[degi]),3))+
' for degree '+str(deg)+'.')
t,p = stats.ttest_rel(sse[0],sse[1])
print(t,p)
# Test for interaction between aHPC and pHPC correlation with ev_conv:
result = sm2.ols(formula='Exact_Age ~ correlation_aHPC + correlation_pHPC + correlation_aHPC * correlation_pHPC', data=tempdf).fit()
print(result.summary()) #no interaction!
for HPC in ['HPC','aHPC','pHPC']:
OLS_model = sm.OLS(tempsize['correlation_'+HPC],tempsize[HPC]).fit() # training the model
residual_values = OLS_model.resid # residual values
r3,p3 = stats.pearsonr(tempsize['Exact_Age'],tempsize['correlation_'+HPC])
X = sm.add_constant(tempsize[['Exact_Age',HPC]]) # adding a constant
model = sm.OLS(tempsize['correlation_'+HPC], X).fit()
predictions = model.predict(X)
print_model = model.summary()
print(print_model)
for HPC in ['HPC','aHPC','pHPC']:
r,p = stats.pearsonr(tempdf['Exact_Age'],tempdf['correlation_'+HPC])
sns.set(font_scale = 2,style="ticks")
fig,ax=plt.subplots(figsize=(7,5))
sns.scatterplot(x='Exact_Age',y="correlation_"+HPC,hue='Age', data=tempdf,palette=colors_age,ax=ax,linewidth=0,alpha = 0.7,legend=False,s=50)
sns.regplot(x='Exact_Age', y="correlation_"+HPC, data=tempdf, scatter=False, ax=ax,color=colors_age[2])
ax.set_xlabel('Age')
ax.set_ylabel('Hippocampus-to-event\ncorrelation')
plt.rcParams['axes.xmargin'] = 0
print(HPC,', r = '+str(np.round(r,2)),', p = '+str(np.round(p,5)))
fig.savefig(figurepath+'HPC/'+HPC+'_Age_vs_bump_xcorr_ev_conv.png', bbox_inches='tight', dpi=300)
for HPC in ['HPC','aHPC','pHPC']:
# no correlation between size and event-response!!
r2,p2 = stats.pearsonr(tempsize[HPC],tempsize['correlation_'+HPC])
# plot indicating size
fig,ax=plt.subplots(figsize=(7,5))
sns.scatterplot(x='Exact_Age', y="correlation_"+HPC, hue=HPC, size=HPC,linewidth=0,alpha=.5,
data=tempsize, ax=ax)
ax.set_xlabel('Age')
ax.set_ylabel('Hippocampus-to-event\ncorrelation')
plt.rcParams['axes.xmargin'] = 0
ax.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)
print(HPC,', r = '+str(np.round(r2,2)),', p = '+str(np.round(p2,5)))
fig.savefig(figurepath+'HPC/'+HPC+'_Age_vs_bump_size.png', bbox_inches='tight', dpi=300)
for HPC in ['HPC','aHPC','pHPC']:
# plot all ages together
sns.set_palette(colors_age)
fig,ax = plt.subplots(1,1,figsize=(7,7))
g = sns.lineplot(x='Time lag [s]', y='correlation_'+HPC, ax=ax, data=dfbumplag, ci=95,color=colors_age[3])
ax.set_xlim([-10,10])
ax.set_xlabel('Time (seconds)')
ax.set_ylabel('Hippocampus-to-event\ncorrelation '+HPC)
ax.margins(x=0)
plt.savefig(figurepath+'HPC/'+HPC+'_bump_all.png', bbox_inches='tight',dpi=300)
for HPC in ['HPC','aHPC','pHPC']:
sns.set(font_scale = 4,style="ticks")
sns.set_palette(colors_age)
# plot timecourse of xcorr
fig,ax = plt.subplots(1,1,figsize=(10,12))
g = sns.lineplot(x='Time lag [s]', y='correlation_'+HPC, hue='Age', ax=ax, data=dfbumplag, ci=95,linewidth = 5)
#ax.set_xlim([-10,10])
ax.set_xlabel('Time (seconds)')
ax.set_ylabel('Hippocampus-to-event\ncorrelation')
leg = ax.legend(loc='center', bbox_to_anchor=(0.5, -0.5))
for line in leg.get_lines():
line.set_linewidth(10)
ax.margins(x=0)
plt.savefig(figurepath+'HPC/'+HPC+'_bump_xcorr_ev_conv.png', bbox_inches='tight',dpi=300)