-
Notifications
You must be signed in to change notification settings - Fork 1
/
colab_notebook_elife.py
192 lines (153 loc) · 7.3 KB
/
colab_notebook_elife.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
# -*- coding: utf-8 -*-
"""NasmythNotebookELife.ipynb
Automatically generated by Colaboratory.
Original file is located at
https://colab.research.google.com/drive/1L65JXwf1JC4S9MtX7YNJti8w69n56Jgx
"""
wt_url = 'https://ftp.ncbi.nlm.nih.gov/geo/samples/GSM5692nnn/GSM5692829/suppl/GSM5692829_mes_wt.hic'
mcph1_url = 'https://ftp.ncbi.nlm.nih.gov/geo/samples/GSM5692nnn/GSM5692830/suppl/GSM5692830_mes_mcph1.hic'
!pip install hic-straw
import straw as straw
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.colors import LinearSegmentedColormap
import pandas as pd
from numpy import linalg as lin_alg
from scipy.sparse import coo_matrix
from scipy import stats
def grab_expected_vector(filepath: str, chrom1: str, resolution: int, norm: str):
footer = straw.getNormExpVectors(filepath, chrom1, chrom1, "oe", norm, "BP", resolution)
return footer.expectedValues
wt_expected = grab_expected_vector(wt_url, '1', 100000, 'KR')
mcph1_expected = grab_expected_vector(mcph1_url, '1', 100000, 'KR')
def normalize(y):
y = np.asarray(y)
y = y/y[0]
return y
wt_expected = normalize(wt_expected)
mcph1_expected = normalize(mcph1_expected)
# 1 = 1MB, 0.1 = 100KB for x axis scaling
def plot_expected_contact_vs_dist(wt_expected, mcph1_expected):
xd = np.asarray(np.linspace(0, len(wt_expected)-1, len(wt_expected)))*.1
#plt.rc('font', size=40)
#plt.rc('axes', titlesize=100)
#plt.rc('axes', labelsize=60)
#plt.rc('xtick', labelsize=60)
#plt.rc('ytick', labelsize=60)
#plt.rc('legend', fontsize=80)
#plt.rc('figure', titlesize=100)
fig, ax = plt.subplots(figsize=(10,10))
plt.semilogy(xd, wt_expected, 'b-', label='WT', linewidth=1)
plt.semilogy(xd, mcph1_expected, 'r-', label='MCPH1', linewidth=1)
plt.ylabel('Average Contacts/Max Contacts')
plt.xlabel('Distance (MB)')
ax.legend()
#ax.legend(prop={'size': 100})
plt.show()
plt.savefig('log_contact_vs_dist.png', dpi=300, bbox_inches='tight')
plt.close()
plot_expected_contact_vs_dist(wt_expected, mcph1_expected)
REDMAP = LinearSegmentedColormap.from_list("bright_red", [(1, 1, 1), (1, 0, 0)])
BLUEREDMAP = LinearSegmentedColormap.from_list("bright_red", [(0, 0, 1), (0, 0, 0), (1, 0, 0)])
def plot_map_neg_pos(dense_matrix, size, outfile):
fig, ax = plt.subplots(figsize=size)
maxval = np.mean(np.abs(dense_matrix))
im = ax.matshow(dense_matrix, cmap=BLUEREDMAP, vmin=-maxval, vmax=maxval)
fig.colorbar(im)
plt.savefig(outfile, dpi=300, bbox_inches='tight')
plt.show()
plt.close()
def grab_hic_region(hic_file, chrom1, resolution, norm, mtype="observed"):
result = straw.straw(mtype, norm, hic_file, str(chrom1), str(chrom1), 'BP', resolution)
row_indices, col_indices, data = list(), list(), list()
for record in result:
row_indices.append(record.binX)
col_indices.append(record.binY)
data.append(record.counts)
if record.binX != record.binY:
row_indices.append(record.binY)
col_indices.append(record.binX)
data.append(record.counts)
row_indices = (np.asarray(row_indices)) / resolution
col_indices = (np.asarray(col_indices)) / resolution
width = int(np.max(row_indices) + 1)
matrix = coo_matrix((data, (row_indices.astype(int), col_indices.astype(int))),
shape=(width, width)).toarray()
matrix[np.isnan(matrix)] = 0
matrix[np.isinf(matrix)] = 0
return matrix
def slice_matrix(matrix, indices):
matrix = matrix[indices, :]
return matrix[:, indices]
def slice_all(z1, z2, r1, r2, indices):
z1 = slice_matrix(z1, indices)
z2 = slice_matrix(z2, indices)
r1 = slice_matrix(r1, indices)
r2 = slice_matrix(r2, indices)
return z1, z2, r1, r2
def paired_remove_sparse_rows(z1, z2, r1, r2):
indices = np.array(np.where(np.sum(z1 > 0, axis=1) > .5 * np.shape(z1)[1])[0])
z1, z2, r1, r2 = slice_all(z1, z2, r1, r2, indices)
indices = np.array(np.where(np.sum(z2 > 0, axis=1) > .5 * np.shape(z2)[1])[0])
return slice_all(z1, z2, r1, r2, indices)
def get_corr_matrix(z):
d_z = pd.DataFrame(z)
d_z = d_z.corr()
np.fill_diagonal(d_z.values, 0.0)
d_z[np.isinf(d_z)] = 0
d_z[np.isnan(d_z)] = 0
return d_z
def get_sorted_vector(vector):
sorted_vector = list(vector)
sorted_vector.sort(key=lambda s: float(s)) # reverse=True
return sorted_vector
def non_zero_median(matrix, indx1, indx2):
temp = matrix[indx1, :][:, indx2].copy()
# return np.mean(temp[temp>0])
return stats.gmean(temp[temp > 0])
def non_zero_median2(temp):
return stats.gmean(temp[temp > 0])
def get_all_nonzero_medians(oe_c1, pos_indices, neg_indices):
aa = non_zero_median(oe_c1, pos_indices, pos_indices)
ab = non_zero_median(oe_c1, pos_indices, neg_indices)
bb = non_zero_median(oe_c1, neg_indices, neg_indices)
return aa, ab, bb, np.sqrt(aa * bb) / ab
def get_eig_and_saddle(oe_c):
corr_matrix = get_corr_matrix(oe_c)
eig_vals, eig_vecs = lin_alg.eig(corr_matrix)
primary_eig = np.squeeze(eig_vecs[:, 0])
primary_eig = primary_eig / np.max(np.abs(primary_eig))
sort_eig_vecs = get_sorted_vector(primary_eig)
saddle = np.outer(sort_eig_vecs, sort_eig_vecs)
return primary_eig, saddle, corr_matrix
def run_ab_enrichment_analysis(hic_file1, hic_file2, chrom1, resolution, norm, mtype):
raw_c1 = grab_hic_region(hic_file1, chrom1, resolution, norm)
raw_c2 = grab_hic_region(hic_file2, chrom1, resolution, norm)
oe_c1 = grab_hic_region(hic_file1, chrom1, resolution, norm, mtype)
oe_c2 = grab_hic_region(hic_file2, chrom1, resolution, norm, mtype)
oe_c1, oe_c2, raw_c1, raw_c2 = paired_remove_sparse_rows(oe_c1, oe_c2, raw_c1, raw_c2)
primary_eig1, saddle1, corr1 = get_eig_and_saddle(oe_c1)
primary_eig2, saddle2, corr2 = get_eig_and_saddle(oe_c2)
pos_cutoff1 = np.median(primary_eig1[primary_eig1 > 0])
neg_cutoff1 = np.median(primary_eig1[primary_eig1 < 0])
max_pos_indices = np.array(np.where((primary_eig1 > pos_cutoff1))[0])
max_neg_indices = np.array(np.where((primary_eig1 < neg_cutoff1))[0])
oe_indices1 = np.argsort(primary_eig1)
oe_indices2 = np.argsort(primary_eig1)
corr_ordered_c1 = np.asarray(corr1)[oe_indices1, :][:, oe_indices1]
corr_ordered_c2 = np.asarray(corr2)[oe_indices2, :][:, oe_indices2]
saddle3 = np.triu(corr_ordered_c2, k=1) + np.tril(corr_ordered_c1, k=-1)
plot_map_neg_pos(saddle3, (10, 10), 'corr_vs_oe_' + str(chrom1) + "_" +
str(resolution) + ".png")
aa, ab, bb, zm1 = get_all_nonzero_medians(oe_c1, max_pos_indices, max_neg_indices)
print("max_aa_C1", aa, "max_ab_C1", ab, "max_bb_C1", bb, "max_enrichment_C1", np.sqrt(aa * bb) / ab)
aa, ab, bb, zm2 = get_all_nonzero_medians(oe_c2, max_pos_indices, max_neg_indices)
print("max_aa_C2", aa, "max_ab_C2", ab, "max_bb_C2", bb, "max_enrichment_C2", np.sqrt(aa * bb) / ab)
run_ab_enrichment_analysis(wt_url, mcph1_url, '11', 250000, 'KR', 'oe')
# for the paper, we used 250kb eigenvectors and used chr11
# but you can also run this at higher resolutions (e.g. 100kb)
# or for different chromosomes
# sometimes, the eigenvector on larger chromosomes (e.g. chr1) can correlate to
# the long vs short arms instead of the AB-compartmentalization signal
# this is something that should be checked in Juicebox before running
# run_ab_enrichment_analysis(wt_url, mcph1_url, '11', 100000, 'KR', 'oe')