-
Notifications
You must be signed in to change notification settings - Fork 8
/
Copy pathpower_stats_simulation.py
189 lines (137 loc) · 7.68 KB
/
power_stats_simulation.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
#!/usr/bin/env python
# coding: utf-8
# In[5]:
import matplotlib
matplotlib.use('Agg')
import pandas as pd
import numpy as np
from scipy.stats import chi2_contingency
import os
from collections import deque
import matplotlib.pyplot as plt
import math
import seaborn as sns
import argparse
import warnings
import sys
def my_get_group(g, key):
if key in g.groups: return True
return False
def produce_power_samples(transcript_isoform_max, total_coverage, n_samples, output):
for num_of_transcripts in range(1, transcript_isoform_max):
d = deque()
for sqrt_coverage in range(2, math.ceil(np.sqrt(total_coverage)) + 1):
coverage = sqrt_coverage ** 2
if ((coverage % 2) == 1):
coverage = coverage + 1
coverage_eR = int(coverage / 2)
coverage_eA = coverage - coverage_eR
delta = np.log2(coverage_eA / coverage_eR)
# produce N samples
for samples in range(1, n_samples):
p_i = np.random.random(num_of_transcripts)
p_i /= p_i.sum()
q_i = np.random.random(num_of_transcripts)
q_i /= q_i.sum()
obs_prob = [p_i, q_i]
chi2_prob, p_prob, dof_prob, expected_prob = chi2_contingency(obs_prob, correction=False)
cohen = round(np.sqrt(chi2_prob / transcript_isoform_max), 1)
e_Ri = np.array(np.random.multinomial(coverage_eR, p_i, 1), dtype=float)
e_Ai = np.array(np.random.multinomial(coverage_eA, q_i, 1), dtype=float)
for j in range(0, len(e_Ri[0])):
if (e_Ri[0][j] == 0 and e_Ai[0][j] == 0):
e_Ri[0][j] = sys.float_info.epsilon
e_Ai[0][j] = sys.float_info.epsilon
obs = [e_Ri, e_Ai]
chi2, p, dof, expected = chi2_contingency(obs)
d.append([coverage, coverage_eR, coverage_eA, cohen, p, delta])
input_stats = pd.DataFrame({'total_count': [item[0] for item in d], 'ref_coverage': [item[1] for item in d],
'alt_coverage': [item[2] for item in d], 'cohen_w': [item[3] for item in d],
'p_value': [item[4] for item in d], 'delta_log2': [item[5] for item in d]})
###########################################################################################################################
subset_input_stats = input_stats.groupby(['cohen_w', 'total_count'])
dist_values = [0, .1, .2, .3, .4, .5, .6, .7, .8, .9, 1] # input_stats.cohen_w.unique()
coverage_values = input_stats.total_count.unique()
data = deque()
for dist in dist_values:
for count in coverage_values:
if (my_get_group(subset_input_stats, (dist, count)) == False):
data.append([dist, count, np.nan])
else:
if (len(subset_input_stats.get_group((dist, count))) <= 1):
print('warning: not enough samples for ' + ' effect size ' + str(
dist) + ' and total coverage ' + str(count) + ' with ' + str(
num_of_transcripts) + ' transcript isoforms')
data.append([dist, count, np.nan])
else:
rejected_null_hypothesis = subset_input_stats.get_group((dist, count))[
(subset_input_stats.get_group((dist, count))['p_value'] < 0.01)]
power = len(rejected_null_hypothesis) / len(subset_input_stats.get_group((dist, count)))
data.append([dist, count, power])
power_df = pd.DataFrame(
{'cohen_w': [item[0] for item in data], 'total_coverage_gene': [item[1] for item in data],
'power': [item[2] for item in data]})
power_df.to_csv(os.path.expanduser(output + '/transcript_isoform_' + str(num_of_transcripts) + '.csv'),
index=False)
def power_plot(power_stats, effect_size, output):
power_combined = pd.read_csv(power_stats)
power_combined_subset = power_combined[(power_combined['cohen_w'] == effect_size)]
power_combined_subset_power = power_combined_subset.groupby(['transcript_isoform', 'total_coverage_gene'])
trans_values = power_combined_subset.transcript_isoform.unique()
trans_values = trans_values[(trans_values > 1)]
coverage_values = power_combined_subset.total_coverage_gene.unique()
data = deque()
for trans in trans_values:
for count in coverage_values:
if (my_get_group(power_combined_subset_power, (trans, count)) == False):
data.append([trans, count, np.nan])
else:
power = power_combined_subset_power.get_group((trans, count))['power'].values[0]
data.append([trans, count, power])
power_df_effectSize = pd.DataFrame(
{'transcript isoforms': [item[0] for item in data], 'total_coverage_gene': [item[1] for item in data],
'power': [item[2] for item in data]})
power_df_effectSize["total_coverage_gene"] = pd.Categorical(power_df_effectSize["total_coverage_gene"],
power_df_effectSize.total_coverage_gene.unique())
data_matrix = power_df_effectSize.pivot("total_coverage_gene", "transcript isoforms", "power")
fig = plt.figure(figsize=(8, 8))
sns.set(font_scale=1.4)
r = sns.heatmap(data_matrix, cmap='BuPu', cbar=False)
r.set_xticklabels(r.get_xticklabels(), rotation=270, horizontalalignment='center')
plt.savefig(os.path.expanduser(output + '/power_' + str(effect_size) + '.pdf'), format='pdf', dpi=3000)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
# REQUIRED
parser.add_argument("--output_path", "--o", required=True, help="Output dir")
parser.add_argument("--isoform_max", type=int, required=True, help="max no. of transcript isoforms")
parser.add_argument("--effect_size", type=float, required=True, help="effect size")
parser.add_argument("--max_coverage", type=int, required=True, choices=range(2, 1024), help="max coverage <= 1024")
parser.add_argument("--iterations", type=int, required=False, default=1000, help="number of simulations")
args = parser.parse_args()
iso_counts = args.isoform_max
output_path = args.output_path
effect_size = args.effect_size
max_coverage = args.max_coverage
num_samples = args.iterations
warnings.filterwarnings("ignore")
script_path = sys.path[0]
print(script_path)
bashCommand = 'mkdir -p ' + output_path
os.system(bashCommand)
bashCommand = 'mkdir ' + output_path + "/temp"
os.system(bashCommand)
bashCommand = 'mkdir ' + output_path + "/temp/sample_files/"
os.system(bashCommand)
bashCommand = 'mkdir ' + output_path + "/plot/"
os.system(bashCommand)
print('produce samples for simuation...')
produce_power_samples(iso_counts + 1, max_coverage + 1, num_samples, output_path + "/temp/sample_files/")
combined_power = pd.DataFrame()
print('combining results...')
for count in range(1, iso_counts + 1):
input_stats = pd.read_csv(output_path + '/temp/sample_files/transcript_isoform_' + str(count) + '.csv')
input_stats['transcript_isoform'] = count
combined_power = combined_power.append(input_stats)
combined_power.to_csv(os.path.expanduser(output_path + '/power_stats_simulation.csv'), index=False)
print('saving plot...')
power_plot(output_path + '/power_stats_simulation.csv', effect_size, output_path + "/plot/")