-
Notifications
You must be signed in to change notification settings - Fork 0
/
predictive_performance_analysis.py
269 lines (230 loc) · 9.98 KB
/
predictive_performance_analysis.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
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
import pandas as pd
import numpy as np
import matplotlib.pylab as plt
import seaborn as sns
import glob
import os
from utils import read_pbz2, makedir
import sys
fig_format = ".png"
ld_scores_colors = {
'D2_0.0': '#E15759',
'D2_0.25': '#C66F6F',
'D2_0.5': '#AC8786',
'D2_0.75': '#919F9C',
'D2_1.0': '#76B7B2',
'R2_0.0': '#B07AA1',
'R2_0.25': '#C17F84',
'R2_0.5': '#D18466',
'R2_0.75': '#E28949',
'R2_1.0': '#F28E2B',
'S-D2_0.0': '#E15759',
'S-D2_0.25': '#C66F6F',
'S-D2_0.5': '#AC8786',
'S-D2_0.75': '#919F9C',
'S-D2_1.0': '#76B7B2',
'S-R2_0.0': '#B07AA1',
'S-R2_0.25': '#C17F84',
'S-R2_0.5': '#D18466',
'S-R2_0.75': '#E28949',
'S-R2_1.0': '#F28E2B'
}
ld_scores_names = {
'D2_0.0': '$\\alpha=0$',
'D2_0.25': '$\\alpha=0.25$',
'D2_0.5': '$\\alpha=0.5$',
'D2_0.75': '$\\alpha=0.75$',
'D2_1.0': '$\\alpha=1$',
'R2_0.0': '$\\alpha=0$',
'R2_0.25': '$\\alpha=0.25$',
'R2_0.5': '$\\alpha=0.5$',
'R2_0.75': '$\\alpha=0.75$',
'R2_1.0': '$\\alpha=1$',
'S-D2_0.0': '$\\alpha=0$',
'S-D2_0.25': '$\\alpha=0.25$',
'S-D2_0.5': '$\\alpha=0.5$',
'S-D2_0.75': '$\\alpha=0.75$',
'S-D2_1.0': '$\\alpha=1$',
'S-R2_0.0': '$\\alpha=0$',
'S-R2_0.25': '$\\alpha=0.25$',
'S-R2_0.5': '$\\alpha=0.5$',
'S-R2_0.75': '$\\alpha=0.75$',
'S-R2_1.0': '$\\alpha=1$',
}
sns.set_context('talk')
plot_global_lines = False
partitioned = True
use_normalized_weights = True
methods = ['D2_0.0', 'D2_0.25', 'D2_0.5', 'D2_0.75', 'D2_1.0']
reference_model = 'D2_1.0'
if partitioned:
methods = ['S-' + m for m in methods]
reference_model = 'S-' + reference_model
# 'Predictive Performance'
performance_category = 'Predictive Performance' #f'Predictive Performance ({reference_model} Weights)'
highly_enriched_cats = [
'Coding_UCSC',
'Conserved_LindbladToh',
'GERP.RSsup4',
'non_synonymous',
'Conserved_Vertebrate_phastCons46way',
'Conserved_Mammal_phastCons46way',
'Conserved_Primate_phastCons46way',
'BivFlnk',
'Ancient_Sequence_Age_Human_Promoter',
'Human_Promoter_Villar_ExAC'
]
metrics = [
'Mean Difference',
'Mean Squared Difference',
'Correlation',
'Weighted Mean Difference',
'Weighted Correlation',
'Weighted Mean Squared Difference',
'Weighted Mean Difference (RWLS Weights)',
'Weighted Correlation (RWLS Weights)',
'Weighted Mean Squared Difference (RWLS Weights)',
f'Weighted Mean Difference ({reference_model} Weights)',
f'Weighted Correlation ({reference_model} Weights)',
f'Weighted Mean Squared Difference ({reference_model} Weights)',
f'Weighted Mean Difference ({reference_model} RWLS Weights)',
f'Weighted Correlation ({reference_model} RWLS Weights)',
f'Weighted Mean Squared Difference ({reference_model} RWLS Weights)'
]
exclude_traits = []
annot_res = []
global_res = []
all_snps_chi2 = []
for trait_file in glob.glob(f"results/regression/EUR/M_5_50_chi2filt/*/*.pbz2"):
trait_res = read_pbz2(trait_file)
trait_name = os.path.basename(os.path.dirname(trait_file))
trait_method = os.path.basename(trait_file).replace('.pbz2', '')
if trait_name in exclude_traits:
continue
for m in methods:
if m == trait_method:
if m == reference_model:
perf_cat = performance_category.split('(')[0].strip()
else:
perf_cat = performance_category
for mbin, mbin_res in trait_res[perf_cat]['Per MAF bin'].items():
for metric in metrics:
if m == reference_model and m in metric:
if 'RWLS' in metric:
score = metric.replace(reference_model + ' ', '')
else:
score = metric.split('(')[0].strip()
else:
score = metric
if use_normalized_weights and 'Weighted' in score:
score = '(Normalized) ' + score
global_res.append({
'Trait': trait_name,
'MAFbin': mbin,
'Metric': metric,
'Method': m,
'Score': mbin_res[score]
})
for metric in metrics + ['Mean Predicted Chisq', 'LRT']:
all_snps_chi2.append({
'Trait': trait_name,
'Metric': metric,
'Method': m,
})
if metric == 'LRT':
all_snps_chi2[-1]['Score'] = trait_res['LRT']
else:
if m == reference_model and m in metric:
if 'RWLS' in metric:
score = metric.replace(reference_model + ' ', '')
else:
score = metric.split('(')[0].strip()
else:
score = metric
all_snps_chi2[-1]['Score'] = trait_res[perf_cat]['Overall'][score]
if partitioned and performance_category == 'Predictive Performance':
for ann, ann_res in trait_res['Annotations'][perf_cat].items():
for mbin, mbin_res in ann_res['Per MAF bin'].items():
for metric in metrics:
if m == reference_model and m in metric:
if 'RWLS' in metric:
score = metric.replace(reference_model + ' ', '')
else:
score = metric.split('(')[0].strip()
else:
score = metric
if use_normalized_weights and 'Weighted' in score:
score = '(Normalized) ' + score
annot_res.append({
'Annotation': ann,
'Trait': trait_name,
'MAFbin': mbin,
'Metric': metric,
'Method': m,
'Score': mbin_res[score]
})
annot_res = pd.DataFrame(annot_res)
global_res = pd.DataFrame(global_res)
all_snps_chi2 = pd.DataFrame(all_snps_chi2)
mean_across_traits = all_snps_chi2.groupby(['Method', 'Metric']).mean()
print(f'Average metrics across all traits and SNP categories:')
print(mean_across_traits)
makedir(f"analysis/{['', 'partitioned/'][partitioned]}{performance_category}/binned")
for metric in metrics:
all_snps_chi2.loc[all_snps_chi2['Metric'] == metric].pivot(
index='Method', columns='Trait', values='Score'
).to_excel(f"analysis/{['', 'partitioned/'][partitioned]}{performance_category}/{metric}.xls")
global_res.loc[global_res['Metric'] == metric].pivot_table(
index=['Method', 'MAFbin'],
columns='Trait',
values='Score'
).to_excel(f"analysis/{['', 'partitioned/'][partitioned]}{performance_category}/binned/{metric}.xls")
print('= = = = = = =')
makedir(f"figures/analysis/{['', 'partitioned/'][partitioned]}{performance_category}/global")
makedir(f"figures/analysis/{['', 'partitioned/'][partitioned]}{performance_category}/annotation")
makedir(f"figures/analysis/{['', 'partitioned/'][partitioned]}{performance_category}/highly_enriched_annotation")
for metric in metrics:
plt.subplots(figsize=(10, 8))
sns.barplot(x='MAFbin', y='Score', hue='Method',
data=global_res.loc[global_res['Metric'] == metric], ci=None,
hue_order=methods,
palette=ld_scores_colors)
plt.xlabel('MAF bin')
plt.ylabel(metric.split("(")[0].strip().replace('Difference', 'Error') + " ($\\chi^2$)")
if plot_global_lines:
for m in np.unique(global_res['Method']):
global_val = mean_across_traits.loc[[(m, metric)], 'Score'].values[0]
plt.axhline(global_val,
ls='--', color=ld_scores_colors[m],
label=f"{m}: {global_val:.3e}")
plt.legend(labels=[ld_scores_names[m] for m in methods])
plt.ylim(-.18, .23)
plt.tight_layout()
plt.savefig(f"figures/analysis/{['', 'partitioned/'][partitioned]}{performance_category}/global/{metric}{fig_format}")
plt.close()
if partitioned and performance_category == 'Predictive Performance':
plt.subplots(figsize=(10, 8))
sns.barplot(x='MAFbin', y='Score', hue='Method',
data=annot_res.loc[annot_res['Metric'] == metric], ci=None,
hue_order=methods,
palette=ld_scores_colors)
plt.xlabel('MAF bin')
plt.ylabel(metric.split("(")[0].strip().replace('Difference', 'Error') + " ($\\chi^2$)")
plt.legend(labels=[ld_scores_names[m] for m in methods])
plt.ylim(-.18, .23)
plt.tight_layout()
plt.savefig(f"figures/analysis/{['', 'partitioned/'][partitioned]}{performance_category}/annotation/{metric}{fig_format}")
plt.close()
plt.subplots(figsize=(10, 8))
sns.barplot(x='MAFbin', y='Score', hue='Method',
hue_order=methods,
data=annot_res.loc[annot_res['Annotation'].isin(highly_enriched_cats) &
(annot_res['Metric'] == metric)], ci=None,
palette=ld_scores_colors)
plt.xlabel('MAF bin')
plt.ylabel(metric.split("(")[0].strip().replace('Difference', 'Error') + " ($\\chi^2$)")
plt.legend(labels=[ld_scores_names[m] for m in methods])
plt.ylim(-.18, .23)
plt.tight_layout()
plt.savefig(f"figures/analysis/{['', 'partitioned/'][partitioned]}{performance_category}/highly_enriched_annotation/{metric}{fig_format}")
plt.close()