SIN-Locator/mainProcess.m at master · PRBioimages/SIN-Locator · 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
% code for "Ge-Wang, Min-Qi Xue, Hong-Bin Shen, and Ying-Ying Xu,...
% Learning multi-view patterns of subcellular localization from protein...
% imaging, sequence and network data"
% Contact: Ying-Ying Xu, yyxu@smu.edu.cn

% initialize
clear
initialize
cType = 'brsvm'; % 'brsvm', 'vote', 'ECOCmkl', 'dnn'

%% datasets and features

% 1. proteins in IF and IHC datasets
load('./data/0_datasetFiles/datasets.mat');

% 2. feature extraction. Features of our dataset have been saved in ./data/7_constructedData
% code can be found at:
% (1) SLF features
%       http://murphylab.web.cmu.edu/software/2012_PLoS_ONE_Reannotation/
%       http://murphylab.web.cmu.edu/software/2008_JProteomeResearch_HPA/
% (2) GenP
%       https://github.com/LorisNanni
% (3) CNNs in Matlab
% (4) Hum-mPLoc 3.0
%       http://www.csbio.sjtu.edu.cn/bioinf/Hum-mPLoc3/
% (5) DeepLoc
%       https://github.com/JJAlmagro/subcellular_localization
% (6) node2vec
%       http://snap.stanford.edu/node2vec/


%% 3. classification
baMethod = 'noba';  % 'noba', 'smote', 'down'
% 3.1 only sequence classification
featTyepSeq = {'IF_seq', 'IHC_seq', 'IF_sequence_deeplocFeats','IHC_sequence_deeplocFeats'};
for i = 1:length(featTyepSeq)
    fType = featTyepSeq{i};  % IF sequence features
    classifyPattern(fType, cType, baMethod, './data/7_constructedData', ...
        './data/8_classification');
end

featTyepSeqCom = {'IF_seq+IF_sequence_deeplocFeats', 'IHC_seq+IHC_sequence_deeplocFeats'};
for i = 1:length(featTyepSeqCom)
    fType = featTyepSeqCom{i};  % IF sequence features
    classifyPattern(fType, cType, 'noba', './data/7_constructedData', ...
        './data/8_classification');
end

%% 3.2 only PPI network classification
% use node2vec features
featTypeNode2vec = getparas('featTypeNode2vec'); % all feat strs
for i = 1:length(featTypeNode2vec)
    fType = featTypeNode2vec{i};
    classifyPattern(fType, cType, 'noba', './data/7_constructedData', ...
        './data/8_classification');
end


%% 3.3 only IF images classification
featTypeIFimg = getparas('featTypeIFimg'); % all feat strs
for i = 1:length(featTypeIFimg)
    classifyPattern(featTypeIFimg{i}, cType, 'noba', './data/7_constructedData',...
        './data/8_classification');
end
% combination of IF features
% The best combination is 'IF_perPro_aveFeats+IF_perPro_bestFitting512PenultimateLayerFeats'
fTypeIFsup = {'IF_perPro_aveFeats+IF_perPro_bestFitting512PenultimateLayerFeats';...

% fTypeIFsup = {'IF_perPro_aveFeats+IF_perPro_aveNanniLF'; ...
%     'IF_perPro_aveFeats+IF_perPro_resnet50Feats';...
%     'IF_perPro_aveFeats+IF_perPro_bestFitting512LastLayerFeats';...
%     'IF_perPro_aveFeats+IF_perPro_bestFitting512PenultimateLayerFeats';
%     'IF_perPro_aveFeats+IF_perPro_bestFitting1024LastLayerFeats';...
%     'IF_perPro_aveFeats+IF_perPro_bestFitting1024PenultimateLayerFeats';...
%     'IF_perPro_aveNanniLF+IF_perPro_bestFitting512LastLayerFeats';...
%     'IF_perPro_aveNanniLF+IF_perPro_bestFitting512PenultimateLayerFeats';...
%     'IF_perPro_aveNanniLF+IF_perPro_bestFitting1024LastLayerFeats';...
%     'IF_perPro_aveNanniLF+IF_perPro_bestFitting1024PenultimateLayerFeats';...
%     'IF_perPro_aveFeats+IF_perPro_aveNanniLF+IF_perPro_bestFitting512LastLayerFeats';...
%     'IF_perPro_aveFeats+IF_perPro_aveNanniLF+IF_perPro_bestFitting512PenultimateLayerFeats';...
%     'IF_perPro_aveFeats+IF_perPro_aveNanniLF+IF_perPro_bestFitting1024LastLayerFeats';...
 %    'IF_perPro_aveFeats+IF_perPro_aveNanniLF+IF_perPro_bestFitting1024PenultimateLayerFeats'...
};
for i = 1:length(fTypeIFsup)
    classifyPattern(fTypeIFsup{i,1}, cType, 'noba', './data/7_constructedData',...
        './data/8_classification');
end

%% 3.4 only IHC images classification
featTypeIHCimg = getparas('featTypeIHCimg'); % all feat strs
% 3.4.1 use whole images
fTypesIHC_wholeImage = featTypeIHCimg(1:11,:);
for i = 1:length(fTypesIHC_wholeImage)
    classifyPattern(fTypesIHC_wholeImage{i}, cType, 'noba', './data/7_construct.edData', ...
        './data/8_classification');
end
% 3.4.2 use patches
% 3.4.2.1 use defined size
sizes = 227;
for i = 1:length(sizes) % for each patch size
    fType = ['IHC_perPro_aveFeats_patchSize' num2str(sizes(i))]; % IHC image aveFeats
    classifyPattern(fType, cType, 'noba', './data/7_constructedData', ...
        './data/8_classification');
    fType = ['IHC_perPro_aveNanniLF_patchSize' num2str(sizes(i))]; % IHC image aveNanniLF
    classifyPattern(fType, cType,  'noba', './data/7_constructedData', ...
        './data/8_classification');
end
% 3.4.2.2 use patch size 227 (CNN)
fTypesIHC_patch = featTypeIHCimg(14:end,:);
for i = 1:length(fTypesIHC_patch)
    classifyPattern(fTypesIHC_patch{i}, cType, 'noba', './data/7_constructedData', ...
        './data/8_classification');
end

% 3.4.3 combination of IF features
% The best combination is 'IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_resnet50Feats_patchSize227'
fTypeIHCsup =  {'IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_resnet50Feats_patchSize227'};
% fTypeIHCsup =
% {'IHC_perPro_aveFeats_wholeImage+IHC_perPro_aveNanniLF_wholeImage';  %run
%     'IHC_perPro_aveFeats_wholeImage+IHC_perPro_resnet50Feats_patchSize227';...  %run
%     'IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_resnet50Feats_patchSize227';...  %run
%     'IHC_perPro_aveFeats_wholeImage+IHC_perPro_resnet50Feats_wholeImage';...  %run
%     'IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_resnet50Feats_wholeImage';...  %run
%     'IHC_perPro_aveFeats_wholeImage+IHC_perPro_AlexnetFeats_patchSize227';...
%     'IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_AlexnetFeats_patchSize227';...
%     'IHC_perPro_aveFeats_wholeImage+IHC_perPro_resnet18Feats_patchSize227';...
%     'IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_resnet18Feats_patchSize227';...
%     'IHC_perPro_aveFeats_wholeImage+IHC_perPro_inceptionv3netFeats_patchSize227';...
%     'IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_inceptionv3netFeats_patchSize227';...
%     'IHC_perPro_aveFeats_wholeImage+IHC_perPro_inceptionresnetv2netFeats_patchSize227';...
%     'IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_inceptionresnetv2netFeats_patchSize227';...
%     'IHC_perPro_aveFeats_wholeImage+IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_resnet50Feats_patchSize227';...
%     'IHC_perPro_aveFeats_wholeImage+IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_AlexnetFeats_patchSize227';...
%     'IHC_perPro_aveFeats_wholeImage+IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_resnet18Feats_patchSize227';...
%     'IHC_perPro_aveFeats_wholeImage+IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_inceptionv3netFeats_patchSize227';...
%     'IHC_perPro_aveFeats_wholeImage+IHC_perPro_aveNanniLF_wholeImage+IHC_perPro_inceptionresnetv2netFeats_patchSize227'};
for i = 1:length(fTypeIHCsup)
    classifyPattern(fTypeIHCsup{i,1}, cType, 'noba', './data/7_constructedData',...
        './data/8_classification');
end

%% 3.5 use three data types
fTypeFinal = {'IF_seqIF_sequence_deeplocFeats+IF_perPro_aveFeatsIF_perPro_bestFitting512PenultimateLayerFeats';...  % seq+img
     'IF_seqIF_sequence_deeplocFeats+IF_perPro_node2vec500wFeats';... % seq+PPI
     'IF_perPro_aveFeatsIF_perPro_bestFitting512PenultimateLayerFeats+IF_perPro_node2vec500wFeats';... % img+PPI
    'IF_seqIF_sequence_deeplocFeats+IF_perPro_aveFeatsIF_perPro_bestFitting512PenultimateLayerFeats+IF_perPro_node2vec500wFeats';... % seq+img+PPI
    'IHC_seqIHC_sequence_deeplocFeats+IHC_perPro_aveNanniLF_wholeImageIHC_perPro_resnet50Feats_patchSize227';... % seq+img
    'IHC_seqIHC_sequence_deeplocFeats+IHC_perPro_node2vec500wFeats';... % seq+PPI
    'IHC_perPro_aveNanniLF_wholeImageIHC_perPro_resnet50Feats_patchSize227+IHC_perPro_node2vec500wFeats';... % img+PPI
    'IHC_seqIHC_sequence_deeplocFeats+IHC_perPro_aveNanniLF_wholeImageIHC_perPro_resnet50Feats_patchSize227+IHC_perPro_node2vec500wFeats'};  % seq+img+PPI
for i = 1:length(fTypeFinal)
    classifyPattern(fTypeFinal{i,1}, cType, 'noba', './data/7_constructedData',...
            './data/8_classification');
end