-
Notifications
You must be signed in to change notification settings - Fork 0
/
process_dna_fish.py
258 lines (218 loc) · 9.83 KB
/
process_dna_fish.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
from plotting_functions import init_subplots
import matplotlib.pyplot as plt
from scipy.ndimage import label
from skimage.feature import peak_local_max
from skimage.measure import find_contours
import numpy as np
from skimage.segmentation import watershed
from skimage.feature import peak_local_max
from scipy import ndimage as ndi
print("Imported all modules")
import pandas as pd
from skimage import measure
from skimage.filters import difference_of_gaussians as dog
import tifffile as tiff
class FishExperiment():
def __init__(self, path_to_dapi, path_to_fluorophore_dict, lowco = 800, highco = 3000, threshold_rel = 0.5,
segment_co = 40, filter_ctla4=False):
self.threshold_rel = threshold_rel
self.lowco = lowco
self.highco = highco
self.dapi_img = tiff.imread(path_to_dapi)
self.dapi_array = np.array(self.dapi_img).sum(axis=2)
self.flurophore_dict = {}
for key, value in path_to_fluorophore_dict.items():
img = tiff.imread(value)
array = np.array(img).sum(axis=2)
setattr(self, key + '_array', array)
self.flurophore_dict[key] = array
if filter_ctla4:
if key == 'Ctla4':
self.flurophore_dict[key] = self.flurophore_dict[key] - self.flurophore_dict[key].min()
bad_inds = self.flurophore_dict[key] > 40_000
struct2 = ndi.generate_binary_structure(2, 2)
dilated_structure = ndi.morphology.binary_dilation(bad_inds, structure=struct2, iterations=30)
self.flurophore_dict[key][dilated_structure] = 0
self.cell_labels = self.segment_cells(co=segment_co)
self.cell_labels = self.permute_labels()
self.all_original_cells = self.cell_labels.copy()
self.cell_ids, self.cell_counts = self.set_label_counts()
self.cell_labels = self.remove_outlier_cells()
self.peak_dict, self.dogmat_dict = self.call_peaks(threshold_rel=threshold_rel)
self.peak_cell_df = self.make_peak_cell_df()
self.valid_cell_ids = np.unique(self.cell_labels[~np.isnan(self.cell_labels)])
def remove_outlier_cells(self, ):
small_cells_idx = (self.cell_counts < self.lowco)
large_cells_idx = (self.cell_counts > self.highco)
good_size_idx = (~small_cells_idx) & (~large_cells_idx)
good_idx = np.where(good_size_idx)[0]
filtered_array = self.cell_labels.copy()
filtered_array = filtered_array.copy()
self.small_cells = self.cell_labels.copy()
self.large_cells = self.cell_labels.copy()
filtered_array[~np.isin(filtered_array, self.cell_ids[good_idx])] = np.nan
self.small_cells[~np.isin(self.cell_labels, self.cell_ids[small_cells_idx])] = np.nan
self.large_cells[~np.isin(self.cell_labels, self.cell_ids[large_cells_idx])] = np.nan
return filtered_array
def segment_cells(self, co=40):
image = self.dapi_array > co
distance = ndi.distance_transform_edt(image)
coords = peak_local_max(distance, footprint=np.ones((3, 3)), labels=image, min_distance=10)
mask = np.zeros(distance.shape, dtype=bool)
mask[tuple(coords.T)] = True
markers, _ = ndi.label(mask)
cell_labels = watershed(-distance, markers, mask=image).astype(float)
return cell_labels
def call_peaks(self, **kwargs):
peaks = {}
dogmats = {}
for key, value in self.flurophore_dict.items():
dogmat, (row, col) = get_peaks(value, **kwargs)
peaks[key] = (row, col)
dogmats[key] = dogmat
return peaks, dogmats
def make_peak_cell_df(self):
rowdata = []
for key, (rows, cols) in self.peak_dict.items():
cell_labels = self.cell_labels[rows, cols]
rowdata.extend(zip([key]*len(rows), cell_labels, rows, cols))
peak_cell_df = pd.DataFrame(rowdata, columns=['fluorophore', 'cell_label', 'row', 'col'])
return peak_cell_df
def set_label_counts(self):
values, counts = np.unique(self.cell_labels, return_counts=True)
values, counts = values[:-1], counts[:-1]
return values, counts
def permute_labels(self):
unique_values = np.unique(self.cell_labels)
permuted_values = np.random.permutation(unique_values)
value_mapping = {original: permuted for original, permuted in zip(unique_values, permuted_values)}
value_mapping[np.nan] = np.nan
permuted_array = np.vectorize(value_mapping.get)(self.cell_labels)
permuted_array[self.cell_labels==0] = np.nan
return permuted_array
def make_qc_plots(n = 5):
pass
# for i
def plot_cell(self, cell_id):
X, Y = np.where(self.cell_labels == cell_id)
sl1, sl2 = slice(min(X), max(X)), slice(min(Y), max(Y))
self.make_qc_plot(sl1, sl2)
def make_qc_plot(self, sl1, sl2, fgsz=(2, 2)):
s1, e1, s2, e2 = sl1.start, sl1.stop, sl2.start, sl2.stop
fig, axs = init_subplots(5, 1, space=.2, fgsz=(2, 2))
axs[0].matshow(self.dapi_array[sl1, sl2], cmap='coolwarm', extent=[s2, e2, e1, s1])
axs[1].matshow(self.all_original_cells[sl1, sl2], cmap='tab20', extent=[s2, e2, e1, s1])
axs[2].matshow(self.small_cells[sl1, sl2], cmap='tab20', extent=[s2, e2, e1, s1])
axs[3].matshow(self.large_cells[sl1, sl2], cmap='tab20', extent=[s2, e2, e1, s1])
axs[4].matshow(self.cell_labels[sl1, sl2], cmap='tab20', extent=[s2, e2, e1, s1])
axs[0].set_title("DAPI")
axs[1].set_title("All Nuclei")
axs[2].set_title("Too Small Nuclei")
axs[3].set_title("Too Large Nuclei")
axs[4].set_title("Valid Nuclei")
for ax in axs:
ax.axhline((s1+e1)//2, linestyle='--', color='lightgray')
ax.axvline((s2+e2)//2, linestyle='--', color='lightgray')
ax.set_xticks([])
ax.set_yticks([])
tmp2 = self.cell_labels[sl1, sl2].copy()
tmp2[np.isnan(tmp2)] = 0
num_features = len(np.unique(tmp2))
us = np.unique(tmp2)
for region_label in range(1, num_features):
region = (tmp2 == us[region_label]).astype(int)
contours = find_contours(region, 0.5)
for contour in contours:
for ax in axs:
ax.plot(contour[:, 1]+s2, contour[:, 0]+s1, linewidth=1, color='black')
n = len(self.flurophore_dict)
fig, axs = init_subplots(2*n, 2, space=.4, fgsz=fgsz)
for c, (name, array) in enumerate(self.flurophore_dict.items()):
axs[c].matshow(array[sl1, sl2], cmap='coolwarm', extent=[s2, e2, e1, s1])
axs[c].set_title(f'{name} (raw)')
peaks_y, peaks_x = self.peak_dict[name]
good_peaks = (peaks_y < e1) & (peaks_y > s1) & (peaks_x < e2) & (peaks_x > s2)
peaks_y, peaks_x = peaks_y[good_peaks], peaks_x[good_peaks]
dogmat = self.dogmat_dict[name]
axs[c+n].matshow(dogmat[sl1, sl2], cmap='coolwarm', extent=[s2, e2, e1, s1])
axs[c+n].scatter(peaks_x, peaks_y-.5, edgecolor='black', facecolor='none', marker='s', s=200, zorder=3)
axs[c+n].set_title(f'{name} (DOG)')
for ax in axs:
ax.axhline((s1+e1)//2, linestyle='--', color='lightgray')
ax.axvline((s2+e2)//2, linestyle='--', color='lightgray')
ax.set_xticks([s2, e2])
ax.set_yticks([s1, e1])
for region_label in range(1, num_features):
region = (tmp2 == us[region_label]).astype(int)
contours = find_contours(region, 0.5)
for contour in contours:
for ax in axs:
ax.plot(contour[:, 1]+s2, contour[:, 0]+s1, linewidth=1, color='black')
def get_peaks(mat,threshold_rel=.5):
dogmat = dog(mat, 1, 2)
maxes = peak_local_max(dogmat, min_distance=4, threshold_rel=threshold_rel)
if len(maxes) > 0:
row, col = list(zip(*maxes))
row, col = list(map(np.array, [row, col]))
else:
row, col = np.array([]), np.array([])
return dogmat, (row, col)
def get_distances(self):
pass
class ControlFishExperiment(FishExperiment):
def __init__(self, path_to_image, lowco = 3200, highco = 30_000, threshold_rel = 0.5):
self.threshold_rel = threshold_rel
self.lowco = lowco
self.highco = highco
self.img = tiff.imread(path_to_image)
self.dapi_array = np.array(self.img)[2]
self.dapi_array = self.dapi_array - np.median(self.dapi_array)
self.dapi_array[self.dapi_array < 0] = 0
self.flurophore_dict = {}
self.flurophore_dict["Arl4c"] = np.array(self.img)[1]
self.flurophore_dict["Ctla4"] = np.array(self.img)[0]
self.cell_labels = self.segment_cells(co=1000)
self.cell_labels = self.permute_labels()
self.all_original_cells = self.cell_labels.copy()
self.cell_ids, self.cell_counts = self.set_label_counts()
self.cell_labels = self.remove_outlier_cells()
self.peak_dict, self.dogmat_dict = self.call_peaks(threshold_rel=threshold_rel)
self.peak_cell_df = self.make_peak_cell_df()
def get_distance_df(fish_expt):
fluor_neighbors = []
n_cells = len(fish_expt.peak_cell_df['cell_label'].unique())
for cell_label, i in fish_expt.peak_cell_df.groupby('cell_label'):
us = i['fluorophore'].unique()
if (fish_expt.valid_cell_ids == cell_label).sum() == 0:
continue
if cell_label == 0:
continue
if len(us) <= 1:
continue
for u1 in us:
for u2 in us:
if u1 == u2:
continue
idx1 = i['fluorophore'] == u1
idx2 = i['fluorophore'] == u2
if (idx1.sum() != 2) or (idx2.sum() != 2):
continue
sub1 = i.loc[idx1]
sub2 = i.loc[idx2]
for _ in range(2):
if (len(sub1) == 0) or (len(sub2) == 0):
continue
d_row = np.add.outer(sub1['row'].values, -sub2['row'].values)
d_col = np.add.outer(sub1['col'].values, -sub2['col'].values)
dists = np.sqrt(d_row * d_row + d_col * d_col)
ind1, ind2 = np.unravel_index(np.argmin(dists), dists.shape)
nucleus_size = fish_expt.cell_counts[fish_expt.cell_ids == cell_label][0]
row_data = [u1, u2, cell_label, nucleus_size, dists[ind1, ind2]]
fluor_neighbors.append(row_data)
sub1 = sub1.drop(sub1.index[ind1])
sub2 = sub2.drop(sub2.index[ind2])
df = pd.DataFrame(fluor_neighbors, columns = ['from', 'to', 'cell', 'nucleus_size', 'distance']
)
n_cells_with_fluors = len(np.unique(df['cell']))
print('# Cells:', n_cells, '; % with fluor: ', n_cells_with_fluors/n_cells)
return df