single_cell_spatial_image/package_pipeline_multiprocessing.py at main · coffee19850519/single_cell_spatial_image · GitHub

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
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
import os,csv,re,json
import pandas as pd
import numpy as np
import scanpy as sc
import math
# import SpaGCN as spg
import cv2
from skimage import io, color
# from SpaGCN.util import prefilter_specialgenes, prefilter_genes
from pipeline_transform_spaGCN_embedding_to_image import transform_embedding_to_image
from generate_embedding import generate_embedding_sp,generate_embedding_sc,generate_embedding_SEDR,generate_embedding_UMAP
from util import filter_panelgenes
import random, torch
from test import segmentation
from inpaint_images import inpaint
import warnings
import argparse
import glob
from find_category import seg_category_map
from multiprocessing import Pool, cpu_count
from case_study import case_study
warnings.filterwarnings("ignore")


def load_data(h5_path, spatial_path, scale_factor_path):
    # Read in gene expression and spatial location
    adata = sc.read_10x_h5(h5_path)
    spatial_all = pd.read_csv(spatial_path, sep=",", header=None, na_filter=False, index_col=0)
    spatial = spatial_all[spatial_all[1] == 1]
    spatial = spatial.sort_values(by=0)
    assert all(adata.obs.index == spatial.index)
    adata.obs["in_tissue"] = spatial[1]
    adata.obs["array_row"] = spatial[2]
    adata.obs["array_col"] = spatial[3]
    adata.obs["pxl_col_in_fullres"] = spatial[4]
    adata.obs["pxl_row_in_fullres"] = spatial[5]
    adata.obs.index.name = 'barcode'
    adata.var_names_make_unique()

    # Read scale_factor_file
    with open(scale_factor_path) as fp_scaler:
        scaler = json.load(fp_scaler)
    adata.uns["spot_diameter_fullres"] = scaler["spot_diameter_fullres"]
    adata.uns["tissue_hires_scalef"] = scaler["tissue_hires_scalef"]
    adata.uns["fiducial_diameter_fullres"] = scaler["fiducial_diameter_fullres"]
    adata.uns["tissue_lowres_scalef"] = scaler["tissue_lowres_scalef"]

    return adata , spatial_all

def pseduo_images_scGNN(h5_path, spatial_path, scale_factor_path, output_folder,scgnnsp_zdim,scgnnsp_alpha,transform_opt):
    # --------------------------------------------------------------------------------------------------------#
    # -------------------------------load data--------------------------------------------------#

    sample = os.path.split(os.path.split(h5_path)[0])[1]
    adata,spatial_all = load_data(h5_path, spatial_path, scale_factor_path)
    # transform optional
    if transform_opt == 'log':
        sc.pp.log1p(adata)
    elif transform_opt == 'logcpm':
        sc.pp.normalize_total(adata,target_sum=1e4)
        sc.pp.log1p(adata)
    elif transform_opt == 'None':
        transform_opt = 'raw'
    else:
        print('transform optional is log or logcpm or None')

    #  panel gene
    # if panel_gene_path != None:  # case study
    #         gene_list = []
    #         with open(panel_gene_path, 'r') as f:
    #             for line in f:
    #                gene_list.append(line.strip())
    #         filter_panelgenes(adata,gene_list)

    print('load data finish')

    scgnnsp_knn_distanceList = ['euclidean']
    scgnnsp_kList = ['6']
    scgnnsp_bypassAE_List = [True, False]

    scgnnsp_bypassAE = scgnnsp_bypassAE_List[1]
    for scgnnsp_dist in scgnnsp_knn_distanceList:
                for scgnnsp_k in scgnnsp_kList:
                    # --------------------------------------------------------------------------------------------------------#
                     # -------------------------------generate_embedding --------------------------------------------------#
                    image_name =sample+'_scGNN_'+ transform_opt +'_PEalpha' +str(scgnnsp_alpha) +'_zdim'+str(scgnnsp_zdim)
                    embedding = generate_embedding_sc(adata, sample=sample, scgnnsp_dist=scgnnsp_dist,
                                                      scgnnsp_alpha=scgnnsp_alpha, scgnnsp_k=scgnnsp_k,
                                                      scgnnsp_zdim=scgnnsp_zdim, scgnnsp_bypassAE=scgnnsp_bypassAE)
                    # embedding = embedding.detach().numpy()
                    adata.obsm["embedding"] = embedding
                    print('generate embedding finish')
                    os.getcwd()

                    # --------------------------------------------------------------------------------------------------------#
                    #         # --------------------------------transform_embedding_to_image-------------------------------------------------#
                    high_img, low_img = transform_embedding_to_image(adata, image_name, output_folder,
                                                                     img_type='lowres',
                                                                     scale_factor_file=True)  # img_type:lowres,hires,both
                    adata.uns["high_img"] = high_img
                    adata.uns["low_img"] = low_img
                    print('transform embedding to image finish')
    # --------------------------------------------------------------------------------------------------------#
    # --------------------------------inpaint image-------------------------------------------------#
    img_path = output_folder+ "/RGB_images/"
    inpaint_path = inpaint(img_path, sample, adata, spatial_all)
    print('generate pseudo images finish')
    return inpaint_path


# def pseudo_images(h5_path, spatial_path, scale_factor_path, output_folder,method, panel_gene_path, pca_opt,log_opt,normalization_opt):
def pseudo_images(h5_path, spatial_path, scale_factor_path, output_folder,method, panel_gene_path, pca_opt, transform_opt):
        # --------------------------------------------------------------------------------------------------------#
        # -------------------------------load data--------------------------------------------------#
    sample = os.path.split(os.path.split(h5_path)[0])[1]
    # print(sample)
    if method == 'spaGCN':
        adata,spatial_all = load_data(h5_path, spatial_path, scale_factor_path)


        if transform_opt == 'log':
            sc.pp.log1p(adata)
        elif transform_opt == 'logcpm':
            sc.pp.normalize_total(adata,target_sum=1e4)
            sc.pp.log1p(adata)
        elif transform_opt == 'None':
            transform_opt = 'raw'
        else:
            print('transform optional is log or logcpm or None')

        print('load data finish')

        pca_list = [32, 50, 64, 128, 256, 1024]
        res_list = np.arange(0.2, 0.7, 0.05)

        # panel gene
        # pca_opt = True
        if panel_gene_path != None:  # case study
            gene_list = []
            with open(panel_gene_path, 'r') as f:
                for line in f:
                   gene_list.append(line.strip())
            filter_panelgenes(adata,gene_list)
            pca_list = [3]
            res_list = [0.65]
            if pca_opt == False:
                pca_list = [0]
        else:
            # threshold
            threshold = 1.0
            adata.X[adata.X < threshold] = 0
            pca_opt = True


        # optical_img_path = os.path.join(data_path,"spatial/tissue_hires_image.png")
        optical_img_path = None
        for pca in pca_list:
            for res in res_list:
                # --------------------------------------------------------------------------------------------------------#
                # -------------------------------generate_embedding --------------------------------------------------#
                image_name = sample+'_spaGCN_'+ transform_opt +'_pca' +str(pca) +'_res'+str(res)
                print(output_folder)
                print(image_name)
                embedding = generate_embedding_sp(adata,pca=pca, res=res,img_path = optical_img_path, pca_opt=pca_opt)
                embedding = embedding.detach().numpy()
                adata.obsm["embedding"] = embedding
                print('generate embedding finish')

                # --------------------------------------------------------------------------------------------------------#
                # --------------------------------transform_embedding_to_image-------------------------------------------------#
                high_img, low_img = transform_embedding_to_image(adata,image_name,output_folder,img_type='lowres',scale_factor_file=True)  # img_type:lowres,hires,both
                adata.uns["high_img"] = high_img
                adata.uns["low_img"] = low_img
                # adata.uns["img_shape"] = 600
                print('transform embedding to image finish')
        # # --------------------------------------------------------------------------------------------------------#
        # # --------------------------------inpaint image-------------------------------------------------#
        img_path = output_folder + "/RGB_images/"
        inpaint_path = inpaint(img_path, sample, adata, spatial_all)
        print('generate pseudo images finish')
        return inpaint_path

    elif method =='scGNN':
        scgnnsp_PEalphaList = [ '0.1', '0.2', '0.3', '0.5', '1.0', '1.2', '1.5', '2.0']
        scgnnsp_zdimList = ['3', '10',  '16',  '32', '64', '128', '256']

        core_num = cpu_count()
        # print(core_num)
        pool = Pool(core_num - 5)
        for scgnnsp_zdim in scgnnsp_zdimList:
                    for scgnnsp_alpha in scgnnsp_PEalphaList:
                        pool.apply_async(pseduo_images_scGNN, (h5_path, spatial_path, scale_factor_path, output_folder,
                                                               scgnnsp_zdim,scgnnsp_alpha,transform_opt,))
        pool.close()
        pool.join()

    elif method == 'UMAP':
        adata,spatial_all = load_data(h5_path, spatial_path, scale_factor_path)


        if transform_opt == 'log':
            sc.pp.log1p(adata)
        elif transform_opt == 'logcpm':
            sc.pp.normalize_total(adata,target_sum=1e4)
            sc.pp.log1p(adata)
        elif transform_opt == 'None':
            transform_opt = 'raw'
        else:
            print('transform optional is log or logcpm or None')

        print('load data finish')

        pc_num_list = [3,16,32, 50, 64, 128, 256, 1024]
        neighbor_list = [5,10,15,30,50]
        for pc_num in pc_num_list:
           for neighbor in neighbor_list:
                # --------------------------------------------------------------------------------------------------------#
                # -------------------------------generate_embedding --------------------------------------------------#
                image_name = sample +'_UMAP_'+ transform_opt + '_pc_num_'+str(pc_num)+ '_neighnor_'+str(neighbor)
                adata = generate_embedding_UMAP(adata,pc_num,neighbor)
                print('generate embedding finish')
                os.getcwd()

                # --------------------------------------------------------------------------------------------------------#
                # --------------------------------transform_embedding_to_image-------------------------------------------------#
                high_img, low_img = transform_embedding_to_image(adata,image_name,pseudo_image_folder,img_type='lowres',scale_factor_file=True)  # img_type:lowres,hires,both
                adata.uns["high_img"] = high_img
                adata.uns["low_img"] = low_img
                print('transform embedding to image finish')


        # --------------------------------------------------------------------------------------------------------#
        # --------------------------------inpaint image-------------------------------------------------#
        img_path = pseudo_image_folder+"pseudo_images/"
        inpaint_path = inpaint(img_path,adata,spatial_all)
        print('generate pseudo images finish')
        return inpaint_path

    elif method == 'SEDR':
        adata,spatial_all = load_data(h5_path, spatial_path, scale_factor_path)

        if transform_opt == 'log':
            sc.pp.log1p(adata)
        elif transform_opt == 'logcpm':
            sc.pp.normalize_total(adata,target_sum=1e4)
            sc.pp.log1p(adata)
        elif transform_opt == 'None':
            transform_opt = 'raw'
        else:
            print('transform optional is log or logcpm or None')

        print('load data finish')

        k_list = [2,4,6,10,20,30,50]
        cell_feat_dim_list = [100,200,500]
        gcn_w_list = [0.1,1.0,5,10.0]

        for K in k_list:
           for cell_feat_dim in cell_feat_dim_list:
               for gcn_w in gcn_w_list:
                    # --------------------------------------------------------------------------------------------------------#
                    # -------------------------------generate_embedding --------------------------------------------------#
                    image_name = sample  +'_SEDR_'+ transform_opt + '_K_'+str(K)+ '_cell_feat_dim_'+str(cell_feat_dim) + '_gcn_w_'+str(gcn_w)
                    adata = generate_embedding_SEDR(adata,K,cell_feat_dim,gcn_w )
                    adata.obsm["embedding"] = adata.obsm['X_SEDR_umap']
                    print('generate embedding finish')
                    os.getcwd()

                    # --------------------------------------------------------------------------------------------------------#
                    # --------------------------------transform_embedding_to_image-------------------------------------------------#
                    high_img, low_img = transform_embedding_to_image(adata,image_name,pseudo_image_folder,img_type='lowres',scale_factor_file=True)  # img_type:lowres,hires,both
                    adata.uns["high_img"] = high_img
                    adata.uns["low_img"] = low_img
                    print('transform embedding to image finish')
        # --------------------------------------------------------------------------------------------------------#
        # --------------------------------inpaint image-------------------------------------------------#
        img_path = pseudo_image_folder+"pseudo_images/"
        inpaint_path = inpaint(img_path,adata,spatial_all)
        print('generate pseudo images finish')
        return inpaint_path


def segmentation_test(h5_path, spatial_path, scale_factor_path, output_path, method,panel_gene_path,pca_opt,transform_opt,checkpoint, device, k):
    pseudo_images(h5_path, spatial_path, scale_factor_path, output_path, method,panel_gene_path,pca_opt,transform_opt)   # output_folder+ "/pseudo_images/"
    img_path = output_path + "/RGB_images/"
    label_path = None
    adata,spatial_all = load_data(h5_path, spatial_path, scale_factor_path)
    top1_csv_name= segmentation(adata,img_path,label_path,method,checkpoint, device, k)
    return top1_csv_name


def segmentation_category_map(h5_path, spatial_path, scale_factor_path, optical_path, output_path, method, panel_gene_path, pca_opt, transform_opt, checkpoint, device, k):
    optical_img = cv2.imread(optical_path)
    category_map = segmentation_test(h5_path, spatial_path, scale_factor_path, output_path, method, panel_gene_path, pca_opt, transform_opt, checkpoint, device, k)
    seg_category_map(optical_img, category_map, output_path)


def segmentation_evaluation(h5_path, spatial_path, scale_factor_path, output_path, method,label_path, panel_gene_path,pca_opt,transform_opt,checkpoint, device, k):
    pseudo_images(h5_path, spatial_path, scale_factor_path, output_path, method, panel_gene_path,pca_opt,transform_opt)
    img_path =output_path + "/RGB_images/"
    adata,spatial_all = load_data(h5_path, spatial_path, scale_factor_path)
    adata.uns["img_shape"] = 600
    top1_csv_name= segmentation(adata,img_path,label_path,method,checkpoint, device, k)


def case_study_test(h5_path, spatial_path, scale_factor_path, output_path, method,  panel_gene_path , pca_opt, transform_opt,r_tuple,g_tuple,b_tuple):
    img_path = pseudo_images(h5_path, spatial_path, scale_factor_path, output_path, method,  panel_gene_path , pca_opt, transform_opt)   # output_folder+ "/pseudo_images/"
    case_study(img_path,r_tuple,g_tuple,b_tuple)