-
Notifications
You must be signed in to change notification settings - Fork 1
/
Compare_svASE_TF_effects.py
209 lines (180 loc) · 8.75 KB
/
Compare_svASE_TF_effects.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
import pandas as pd
import Utils as ut
import PlotUtils as pu
import numpy as np
from scipy import stats
from GetASEStats import create_latex_command
from numpy import int8, int32, int64
if 'mpl' in locals():
import matplotlib.pyplot as mpl
HAS_MPL = True
else:
HAS_MPL = False
EPSILON = .1
if __name__ == "__main__":
data = {}
expr = pd.read_table('godot/summary_wasp.tsv',
**ut.pd_kwargs).drop(['---'], axis=1)
hyb = expr.select(**ut.sel_startswith(('melXsim', 'simXmel')))
parental = expr.select(**ut.sel_startswith(('melXmel', 'simXsim')))
melXmel = expr.select(**ut.sel_startswith('melXmel')).rename(columns=lambda x:x.split('_sl')[1])
simXsim = expr.select(**ut.sel_startswith('simXsim')).rename(columns=lambda x:x.split('_sl')[1])
higher_parental = melXmel.where(melXmel > simXsim, simXsim)
lower_parental = melXmel.where(melXmel < simXsim, simXsim)
mel_sim_ratio = (melXmel + EPSILON) / (simXsim + EPSILON)
is_mat = pd.read_table('analysis/results/maternal.tsv', squeeze=True,
header=None, **ut.pd_kwargs)
mat_genes = is_mat.index[is_mat]
xs = ut.get_xs(expr)
hyb_xs = ut.get_xs(hyb)
mel_xs = np.linspace(0, 1, 27, endpoint=True)
sim_xs = np.linspace(0, 1, 27, endpoint=True)
parental_xs = ut.get_xs(parental)
kwargs = pu.kwargs_expr_heatmap.copy()
kwargs['progress_bar'] = False
kwargs_with_labels = pu.kwargs_expr_heatmap.copy()
kwargs['draw_row_labels'] = False
kwargs['box_height'] = 1
distro_mels = []
distro_sims = []
target_regions = {
'Kr_ant_bg': ('Kr', (0.33, 0.44)),
'Kr_central': ('Kr', (0.55, 0.72)),
'Kr_post_bg': ('Kr', (0.74, 0.85)),
'hb_ant_tip': ('hb', (0, .10)),
'hb_ant_stripe': ('hb', (.20, .50)),
'hb_interstripe': ('hb', (.60, .75)),
'hb_post_stripe': ('hb', (.80, .90)),
#('prd', (0, .1)),
}
nmegs = [] # Non maternal expressed gene lists
for region_name, (tf, region) in target_regions.items():
peak_genes = {line.strip() for line in
open(
#'analysis/results/{}_1000_peak_genes_2500bp_tss.txt'
'analysis/results/hb_wt_emd_0.1.txt'
#'analysis/results/{}_tss_genes.txt'
.format(tf.lower()))
if line.strip() in expr.index}
data['{}diff'.format(tf)] = len(peak_genes)
peak_genes.add(tf)
all_in_region_ix = (region[0] <= xs) & (xs < region[1])
mel_in_region_ix = (region[0] <= mel_xs) & (mel_xs < region[1])
sim_in_region_ix = (region[0] <= sim_xs) & (sim_xs < region[1])
parental_in_region_ix = (region[0] <= parental_xs) & (parental_xs < region[1])
gene_expr_level = parental.ix[peak_genes, parental_in_region_ix].min(axis=1)
expr_in_region = ut.true_index(gene_expr_level > -1)
non_mat_expr_genes = expr_in_region.difference(mat_genes)
non_mat_expr_genes = mel_sim_ratio.ix[non_mat_expr_genes,
mel_in_region_ix].mean(axis=1).sort_values().index
nmegs.append(non_mat_expr_genes)
pu.svg_heatmap(expr.ix[non_mat_expr_genes, all_in_region_ix],
'analysis/results/non_mat_{}_{:0.2f}-{:0.2f}_expr_genes.svg'
.format(tf, region[0], region[1]),
squeeze_rows=np.nanmean, **kwargs)
mel_in_region = (melXmel.ix[non_mat_expr_genes, mel_in_region_ix]
.divide(parental.ix[non_mat_expr_genes, :]
.max(axis=1), axis=0)
.mean(axis=1))
sim_in_region = (simXsim.ix[non_mat_expr_genes, sim_in_region_ix]
.divide(parental.ix[non_mat_expr_genes, :]
.max(axis=1), axis=0)
.mean(axis=1))
res = stats.ttest_rel( mel_in_region, sim_in_region, nan_policy='omit',)
distro_mels.append(mel_in_region.dropna())
distro_sims.append(sim_in_region.dropna())
cutoff = .25
data['num_{}_change'.format(region_name)] = (
sum(abs(mel_in_region - sim_in_region)
> cutoff))
data['frac_higher_{}'.format(region_name)] = (
sum((mel_in_region - sim_in_region) > cutoff)
/ data['num_{}_change'.format(region_name)]
)
data['prob_higher_{}'.format(region_name)] = stats.binom_test(
[sum((mel_in_region - sim_in_region) > cutoff),
sum((sim_in_region - mel_in_region) > cutoff)]
)
print(region_name, tf, region, data['prob_higher_{}'.format(region_name)],
'mel', sum((mel_in_region - sim_in_region) > cutoff),
'sim', sum((sim_in_region - mel_in_region) > cutoff),)
((mel_in_region - sim_in_region)
.sort_values()
.to_csv('analysis/results/{}_melsim_change.tsv'.format(region_name), sep='\t')
)
pu.svg_heatmap(
expr
.ix[ut.true_index((mel_in_region - sim_in_region) > cutoff)]
.select(**ut.sel_startswith(('melXmel', 'simXsim'))),
'analysis/results/{}_mel_higher.svg'.format(region_name),
**kwargs_with_labels
)
pu.svg_heatmap(
expr
.ix[ut.true_index((mel_in_region - sim_in_region) < -cutoff)]
.select(**ut.sel_startswith(('melXmel', 'simXsim'))),
'analysis/results/{}_sim_higher.svg'.format(region_name),
**kwargs_with_labels
)
x_range = np.arange(0, 1, .05)
if HAS_MPL:
print("Violinning")
mpl.figure()
mpl.violinplot([(dm - ds) for dm, ds in zip(distro_mels, distro_sims)])
mpl.xticks(np.arange(len(target_regions))+1,
[region for region in target_regions],
rotation=90)
mpl.tight_layout()
tfs = [region_name.split('_')[0] for region_name in target_regions]
for tf in tfs:
fig1 = mpl.figure(figsize=(8, 6))
ax1 = mpl.gca()
fig2 = mpl.figure(figsize=(8, 6))
ax2 = mpl.gca()
for nmeg, dm, ds, region_name in zip(
nmegs, distro_mels, distro_sims, target_regions,
):
target, region = target_regions[region_name]
if target.lower() != tf.lower(): continue
region_diffs = dm - ds
control_region = (melXmel.ix[target] < 5) & (simXsim.ix[target] < 5)
control_diffs = ((melXmel.ix[nmeg, control_region]
- simXsim.ix[nmeg, control_region])
.divide(parental.ix[nmeg, :] .max(axis=1), axis=0)
).mean(axis=1)
ax1.semilogy(100*x_range,
[sum(region_diffs > i)/ sum(region_diffs < -i)
for i in x_range],
linewidth=4,
label='{:0.0%} - {:0.0%}'.format( *region), basey=2)
ax2.semilogy(100*x_range,
[
stats.binom_test([sum(region_diffs > i),
sum(region_diffs < -i)],
#p=(sum(control_diffs > i)
#/ (sum(control_diffs > i)
#+ sum(control_diffs < -i))
#)
p=0.5,
)
for i in x_range
],
linewidth=4,
label='{:0.0%} - {:0.0%}'.format( *region),
basey=10)
ax2.invert_yaxis()
ax1.legend(loc='upper left')
ax1.hlines(1, 0, 100)
ax1.set_ylabel(r'$\#(\Delta > x) \div \#(\Delta < -x)$')
ax2.hlines(0.05, 0, 100, label='nominal p=.05')
#ax2.legend(loc='lower left')
ax2.set_ylabel('binomial p value (vs 50%)')
pu.minimize_ink(ax1)
pu.minimize_ink(ax2)
fig1.savefig('analysis/results/{}_bound_Ns.png'.format(tf), dpi=300)
fig2.savefig('analysis/results/{}_bound_ps.png'.format(tf), dpi=300)
with open('analysis/results/tf_stats.tex', 'w') as outf:
for var, val in data.items():
numeric = isinstance(val, (float, int, int64, int32, int8))
frac = var.lower().startswith('frac')
outf.write(create_latex_command(var, val, numeric, frac))