-
Notifications
You must be signed in to change notification settings - Fork 0
/
evaluation.py
155 lines (137 loc) · 5.43 KB
/
evaluation.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
import numpy as np
from scipy.stats import pearsonr
import pandas as pd
from scipy.stats import kruskal
from scipy.sparse import lil_matrix, csr_matrix, vstack
from sys import argv
import os
import matplotlib.pyplot as plt
script, bin_file, startx, starty, patchsize, nucl_seg, cell_seg1, cell_seg2 = argv
verbose = False
def bin2exp():
#find xmin ymin
xall = []
yall = []
df = pd.read_csv(bin_file, sep='\t')
xmin = df['x'].min()
ymin = df['y'].min()
# print(xmin, ymin)
#find all the genes in the range
geneid = {}
genecnt = 0
allgenes = []
with open(bin_file) as fr:
header = fr.readline()
for line in fr:
gene, x, y, count = line.split()
if int(x) - xmin >= int(startx) and int(x) - xmin < int(startx) + int(patchsize) and int(y) - ymin >= int(starty) and int(y) - ymin < int(starty) + int(patchsize):
if gene not in geneid:
geneid[gene] = genecnt
allgenes.append(gene)
genecnt += 1
print(f"total gene number: {genecnt}")
posexp = lil_matrix((int(patchsize) * int(patchsize), genecnt), dtype=np.int8)
with open(bin_file) as fr:
header = fr.readline()
for line in fr:
gene, x, y, count = line.split()
if gene not in geneid:
continue
if int(x) - xmin >= int(startx) and int(x) - xmin < int(startx) + int(patchsize) and int(y) - ymin >= int(starty) and int(y) - ymin < int(starty) + int(patchsize):
idx = int((int(x) - xmin - int(startx)) * int(patchsize) + (int(y) - ymin - int(starty)))
posexp[idx, geneid[gene]] = int(count)
posexp = posexp.tocsr()
return posexp, genecnt, allgenes
watershed_nucleus = {}
with open(nucl_seg) as fr:
for line in fr:
coord, cell = line.split()
if cell not in watershed_nucleus:
watershed_nucleus[cell] = [coord]
else:
watershed_nucleus[cell].append(coord)
seg2_cell = {}
with open(cell_seg2) as fr:
for line in fr:
coord, cell = line.split()
if cell not in seg2_cell:
seg2_cell[cell] = [coord]
else:
seg2_cell[cell].append(coord)
seg1_cell = {}
with open(cell_seg1) as fr:
for line in fr:
coord, cell = line.split()
if cell not in seg1_cell:
seg1_cell[cell] = [coord]
else:
seg1_cell[cell].append(coord)
nucleus2cell1= {}
nucleus2cell2 = {}
for nu in watershed_nucleus:
mapcell = 0
intsc_ze = 0
for cell in seg1_cell:
intsc = set(watershed_nucleus[nu]).intersection(seg1_cell[cell])
if len(intsc) > intsc_ze:
intsc_ze = len(intsc)
mapcell = cell
if mapcell:
nucleus2cell1[nu] = mapcell
for nu in watershed_nucleus:
mapcell = 0
intsc_ze = 0
for cell in seg2_cell:
intsc = set(watershed_nucleus[nu]).intersection(seg2_cell[cell])
if len(intsc) > intsc_ze:
intsc_ze = len(intsc)
mapcell = cell
if mapcell:
nucleus2cell2[nu] = mapcell
posexp, genecnt, allgenes = bin2exp()
allgenes = np.array(allgenes)
corr_seg1 = []
corr_seg2 = []
for nu in nucleus2cell1:
if nu in nucleus2cell2:
int_part = set(seg1_cell[nucleus2cell1[nu]]).intersection(seg2_cell[nucleus2cell2[nu]])
diff_seg1 = set(seg1_cell[nucleus2cell1[nu]]).difference(int_part)
diff_seg2 = set(seg2_cell[nucleus2cell2[nu]]).difference(int_part)
exp_int = np.zeros(genecnt)
exp_seg1 = np.zeros(genecnt)
exp_seg2 = np.zeros(genecnt)
for bin in int_part:
idx = int(bin.split(':')[0]) * int(patchsize) + int(bin.split(':')[1])
exp_int += np.squeeze(posexp[idx].toarray())
for bin in diff_seg1:
idx = int(bin.split(':')[0]) * int(patchsize) + int(bin.split(':')[1])
exp_seg1 += np.squeeze(posexp[idx].toarray())
for bin in diff_seg2:
idx = int(bin.split(':')[0]) * int(patchsize) + int(bin.split(':')[1])
exp_seg2 += np.squeeze(posexp[idx].toarray())
if np.sum(exp_int) >= 100 and np.sum(exp_seg1) >= 100 and np.sum(exp_seg2) >= 100:
r, p = pearsonr(exp_int, exp_seg1)
if np.isnan(r):
corr_seg1.append(0)
else:
corr_seg1.append(r)
r, p = pearsonr(exp_int, exp_seg2)
if np.isnan(r):
corr_seg2.append(0)
else:
corr_seg2.append(r)
if verbose:
print(nu, len(int_part), len(diff_seg1), len(diff_seg2), (nucleus2cell1[nu], len(seg1_cell[nucleus2cell1[nu]])), (nucleus2cell2[nu], len(seg2_cell[nucleus2cell2[nu]])))
print('mean corr_seg1:', np.mean(corr_seg1), 'mean corr_seg2:', np.mean(corr_seg2), 'median corr_seg1:', np.median(corr_seg1), 'median corr_seg2:', np.median(corr_seg2))
if not verbose:
print('mean corr_seg1:', np.mean(corr_seg1), 'mean corr_seg2:', np.mean(corr_seg2), 'median corr_seg1:', np.median(corr_seg1), 'median corr_seg2:', np.median(corr_seg2))
import numpy as np
import matplotlib.pyplot as plt
data = [corr_seg1, corr_seg2]
_, p = kruskal(corr_seg1, corr_seg2)
print(f"Kruskal-Wallis H-test p-value: {p}, Sample num: {len(corr_seg1)}")
fig, ax = plt.subplots()
ax.boxplot(data)
if not os.path.exists('results'):
os.makedirs('results')
plt.savefig('results/r_boxplt_' + startx + ':' + starty + ':' + patchsize + ':' + patchsize + '.png')