-
Notifications
You must be signed in to change notification settings - Fork 3
/
main.m
243 lines (194 loc) · 8.34 KB
/
main.m
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
clc;
close all;
%
R = 5; % Number of independent simulations, for nuclear receptors (4)
% we recommond to set the number of simulation up to 20 or more
% since the performance is not so stable because the network
% is too small
K = 10; % K-fold cross validation
% Initialize parameter alpha
% Adjust these parameters will effect the performances of the method
% The optimal parameters for AUC may differ from those for AUPR
paraT = [0.13 0.12 0.18 0.26 0.1]; % alpha for target similarity A
paraD = [0.23 0.15 0.15 0.25 0.15]; % alpha for drug similarity A^T
paraST= [1.0 1.2 1.4 1.5]; % alpha for target similarity matrix
paraSD= [1.0 1.0 1.0 1.0]; % alpha for drug similarity matrix
data = 1 % Dataset to run, change this value for running different datasets
if data == 1
dg = dlmread('dataset/en_sim_dg.txt');
dc = dlmread('dataset/en_sim_dc.txt');
adj = dlmread('dataset/en_adj.txt');
elseif data == 2
dg = dlmread('dataset/ic_sim_dg.txt');
dc = dlmread('dataset/ic_sim_dc.txt');
adj = dlmread('dataset/ic_adj.txt');
elseif data == 3
dg = dlmread('dataset/gpcr_sim_dg.txt');
dc = dlmread('dataset/gpcr_sim_dc.txt');
adj = dlmread('dataset/gpcr_adj.txt');
elseif data == 4
dg = dlmread('dataset/nr_sim_dg.txt');
dc = dlmread('dataset/nr_sim_dc.txt');
adj = dlmread('dataset/nr_adj.txt');
elseif data == 5
net = dlmread('dataset/matador_adj.txt');
adj = construct_DTI_matrix(net);
end
precision_SD = []; recall_SD = [];
precision_ST = []; recall_ST = [];
precision_AT = []; recall_AT = [];
precision_AD = []; recall_AD = [];
precision_A = []; recall_A = [];
precision_ADT = []; recall_ADT =[];
aucSD = zeros(1,K);
auprSD = zeros(1,K);
aucST = zeros(1,K);
auprST = zeros(1,K);
aucAT = zeros(1,K);
auprAT = zeros(1,K);
aucAD = zeros(1,K);
auprAD = zeros(1,K);
aucA = zeros(1,K);
auprA = zeros(1,K);
aucADT = zeros(1,K);
auprADT = zeros(1,K);
auc_SD = zeros(1,R);
aupr_SD = zeros(1,R);
auc_ST = zeros(1,R);
aupr_ST = zeros(1,R);
auc_AT = zeros(1,R);
aupr_AT = zeros(1,R);
auc_AD = zeros(1,R);
aupr_AD = zeros(1,R);
auc_A = zeros(1,R);
aupr_A = zeros(1,R);
auc_ADT = zeros(1,R);
aupr_ADT = zeros(1,R);
for r = 1 : R % Number of simulations
disp('===========================================================');
disp(['============= r = ' num2str(r) ' ========== of ' num2str(R) ' =======']);
disp('===========================================================');
fprintf('\n');
y = adj;
%[tr te] = divideMatrixCV(adj,K); % Dividing the dataset to k-fold subsets
%crossval_idx = divideMatrixCR(adj,K,2); % Dividing the dataset to k-fold subsets
crossval_idx = crossvalind('Kfold',y(:),K);
for i = 1 : K % each fold at a time
disp(['--- Run ' num2str(r) ' of ' num2str(R) ', k ' num2str(i) ' of ' num2str(K) ' ---']);
train_idx = find(crossval_idx~=i);
test_idx = find(crossval_idx==i);
y_train = y;
y_train(test_idx) = 0;
train = y_train;
test = y - train;
yy=y;
yy(yy==0)=-1;
if data == 5 % Working with Matador
% Note that for Matador, the target similarity obtained from
% the interaction information plays more important role than
% the drug similarity and the even the combination
X_AT = solve_lrr(train',train', paraT(data)); % X^*_AT
Z_AT = (train'*X_AT)';
statsST = evaluate_performance(Z_AT(test_idx),yy(test_idx),'classification');
aucST(i) = statsST.auc;
auprST(i) = statsST.aupr;
X_AD = solve_lrr(train,train, paraD(data)); % X^*_A
Z_AD = train*X_AD;
statsSD = evaluate_performance(Z_AD(test_idx),yy(test_idx),'classification');
aucSD(i) = statsSD.auc;
auprSD(i) = statsSD.aupr;
Z_A = (Z_AT + Z_AD)/2;
statsA = evaluate_performance(Z_A(test_idx),yy(test_idx),'classification');
aucA(i) = statsA.auc;
auprA(i) = statsA.aupr;
else
% Working with drug compound
dc = (dc+dc')/2;
X_SD = solve_lrr(dc,dc, paraSD(data)); % Computing X^*_{SD}
Z_D = train * X_SD; % Projecting A onto X^*_{SD}, Z_D
statsSD = evaluate_performance(Z_D(test_idx),yy(test_idx),'classification');
aucSD(i) = statsSD.auc;
auprSD(i) = statsSD.aupr;
% Working with protein sequence (target)
dg = (dg + dg')/2;
X_ST = solve_lrr(dg,dg, paraST(data)); % Compute X^*_{ST}
Z_T = X_ST * train; % Projecting A onto X^*_{ST}, Z_T
statsST = evaluate_performance(Z_T(test_idx),yy(test_idx),'classification');
aucST(i) = statsST.auc;
auprST(i) = statsST.aupr;
% Working with adjacency matrix (interaction information),
% target similarity
X_AT = solve_lrr(train,train,paraT(data)); % Compute X^*_{AT}
Z_AT = train * X_AT;
statsAT = evaluate_performance(Z_AT(test_idx),yy(test_idx),'classification');
aucAT(i) = statsAT.auc;
auprAT(i) = statsAT.aupr;
% Working with adjacency matrix (interaction information),
% drug similarity
X_AD = solve_lrr(train',train',paraD(data)); % Compute X^*_{AD}
Z_AD = (train' * X_AD)'; % Projecting A on X^*_{D}
statsAD = evaluate_performance(Z_AD(test_idx),yy(test_idx),'classification');
aucAD(i) = statsAD.auc;
auprAD(i) = statsAD.aupr;
% Combine the two Z_AT and Z_AD
Z_A = (Z_AT + Z_AD)/2;
statsA = evaluate_performance(Z_A(test_idx),yy(test_idx),'classification');
aucA(i) = statsA.auc;
auprA(i) = statsA.aupr;
Z_ADT = 0.25*Z_T + 0.25*Z_D + 0.5*Z_A; % Combine all the information
statsADT = evaluate_performance(Z_ADT(test_idx),yy(test_idx),'classification');
aucADT(i) = statsADT.auc;
auprADT(i) = statsADT.aupr;
end
fprintf('\n');
end
fprintf('\n');
if data ~= 5
auc_SD(r) = mean(aucSD);
aupr_SD(r) = mean(auprSD);
auc_ST(r) = mean(aucST);
aupr_ST(r) = mean(auprST);
auc_AT(r) = mean(aucAT);
aupr_AT(r) = mean(auprAT);
auc_AD(r) = mean(aucAD);
aupr_AD(r) = mean(auprAD);
auc_A(r) = mean(aucA);
aupr_A(r) = mean(auprA);
auc_ADT(r) = mean(aucADT);
aupr_ADT(r) = mean(auprADT);
else
auc_AT(r) = mean(aucAT);
aupr_AT(r) = mean(auprAT);
auc_AD(r) = mean(aucAD);
aupr_AD(r) = mean(auprAD);
auc_A(r) = mean(aucA);
aupr_A(r) = mean(auprA);
end
end
%% Print results
if data == 5
disp(['The AUC from Z_ST: ' num2str(auPRAT(auc_AT))]);
disp(['The AUPR from Z_ST: ' num2str(mean(aupr_AT))]);
disp(['The AUC from Z_SD: ' num2str(auPRAT(auc_AD))]);
disp(['The AUPR from Z_SD: ' num2str(mean(aupr_AD))]);
disp(['The AUC from Z_A: ' num2str(auPRAT(auc_A))]);
disp(['The AUPR from Z_A: ' num2str(mean(aupr_A))]);
else
disp(['The AUC from Z_SD: ' num2str(mean(auc_SD))]);
disp(['The AUPR from Z_SD: ' num2str(mean(aupr_SD))]);
fprintf('\n');
disp(['The AUC from Z_ST: ' num2str(mean(auc_ST))]);
disp(['The AUPR from Z_ST: ' num2str(mean(aupr_ST))]);
fprintf('\n');
disp(['The AUC from Z_AD: ' num2str(mean(auc_AT))]);
disp(['The AUPR from Z_AD: ' num2str(mean(aupr_AT))]);
fprintf('\n');
disp(['The AUC from Z_AT: ' num2str(mean(auc_AD))]);
disp(['The AUPR from Z_AT: ' num2str(mean(aupr_AD))]);
fprintf('\n');
disp(['The AUC from Z_A: ' num2str(mean(auc_A))]);
disp(['The AUPR from Z_A: ' num2str(mean(aupr_A))]);
fprintf('\n');
disp(['The AUC from Z_ADT: ' num2str(mean(auc_ADT))]);
disp(['The AUPR from Z_ADT: ' num2str(mean(aupr_ADT))]);
end