Predicting Drugs Side Effects Based on Chemical-Chemical Interactions and Protein-Chemical Interactions
Table 1
The first 20 prediction accuracies of the interaction-based and similarity-based methods in identifying the side effects of drugs in the training and test datasets.
Prediction order
Interaction-based
Similarity-based
Difference
Training dataset
Test dataset
Training dataset
Test dataset
Training dataseta
Test datasetb
1
86.30%
89.16%
83.64%
87.95%
2.66%
1.20%
2
80.45%
83.13%
79.12%
83.13%
1.33%
0.00%
3
77.13%
84.34%
75.00%
79.52%
2.13%
4.82%
4
72.61%
81.93%
71.41%
75.90%
1.20%
6.02%
5
73.40%
77.11%
68.22%
74.70%
5.19%
2.41%
6
68.75%
75.90%
66.89%
71.08%
1.86%
4.82%
7
67.69%
67.47%
64.76%
57.83%
2.93%
9.64%
8
64.23%
65.06%
59.97%
65.06%
4.26%
0.00%
9
63.70%
68.67%
58.78%
57.83%
4.92%
10.84%
10
57.71%
57.83%
57.31%
60.24%
0.40%
−2.41%
11
59.18%
60.24%
56.38%
67.47%
2.79%
−7.23%
12
59.18%
69.88%
56.65%
51.81%
2.53%
18.07%
13
57.31%
61.45%
54.79%
53.01%
2.53%
8.43%
14
55.85%
59.04%
53.86%
62.65%
1.99%
−3.61%
15
54.12%
54.22%
50.66%
57.83%
3.46%
−3.61%
16
53.86%
59.04%
52.66%
55.42%
1.20%
3.61%
17
51.33%
39.76%
50.00%
60.24%
1.33%
−20.48%
18
54.52%
62.65%
51.73%
53.01%
2.79%
9.64%
19
50.00%
56.63%
50.00%
38.55%
0.00%
18.07%
20
47.21%
44.58%
47.74%
51.81%
−0.53%
−7.23%
Percentages in this column were calculated by percentages in column 2 minus percentages in column 4.
bPercentages in this column were calculated by percentages in column 3 minus percentages in column 5.