Breast Cancer Prognosis Risk Estimation Using Integrated Gene Expression and Clinical Data
Table 4
Prognosis based classification results of the IPRE, IPRE (G), IPRE (C), and seven other existing algorithms in the testing dataset of (a) the Desmedt and (b) the Vijver.
Algorithm
TN
FP
TP
FN
SE (95% CI)
SP (95% CI)
ACC (95% CI)
(a) Desmedt dataset
GGI
190
86
66
28
10
0.737 (0.60 to 0.88)
0.566 (0.49 to 0.64)
0.600 (0.53 to 0.67)
70 g
190
42
110
34
4
0.895 (0.80 to 0.99)
0.276 (0.21 to 0.35)
0.400 (0.33 to 0.47)
76 g
190
80
72
23
15
0.605 (0.45 to 0.76)
0.526 (0.45 to 0.61)
0.542 (0.47 to 0.61)
112 g
190
64
88
34
4
0.895 (0.80 to 0.99)
0.421 (0.34 to 0.50)
0.516 (0.45 to 0.59)
IGS
190
90
62
21
17
0.553 (0.39 to 0.71)
0.592 (0.51 to 0.67)
0.584 (0.51 to 0.65)
21 g
190
25
127
38
0
1 (1.00 to 1.00)
0.164 (0.11 to 0.22)
0.332 (0.27 to 0.40)
ITI
190
124
28
12
26
0.316 (0.17 to 0.46)
0.816 (0.75 to 0.88)
0.716 (0.65 to 0.78)
IPRE (C)
190
104
48
9
29
0.237 (0.10 to 0.37)
0.684 (0.61 to 0.76)
0.595 (0.53 to 0.67)
IPRE (G)
190
128
24
19
19
0.500 (0.34 to 0.66)
0.842 (0.78 to 0.90)
0.774 (0.71 to 0.83)
IPRE
190
133
19
23
15
0.605 (0.45 to 0.76)
0.875 (0.82 to 0.93)
0.821 (0.77 to 0.88)
(b) Vijver dataset
GGI
150
68
38
28
16
0.636 (0.49 to 0.78)
0.642 (0.55 to 0.73)
0.640 (0.56 to 0.72)
70 g
150
51
55
39
5
0.886 (0.79 to 0.98)
0.481 (0.39 to 0.58)
0.600 (0.52 to 0.68)
76 g
150
65
41
28
16
0.636 (0.49 to 0.78)
0.613 (0.52 to 0.71)
0.620 (0.54 to 0.70)
112 g
150
42
64
33
11
0.750 (0.62 to 0.88)
0.396 (0.30 to 0.49)
0.500 (0.42 to 0.58)
IGS
150
38
68
40
4
0.909 (0.82 to 0.99)
0.358 (0.27 to 0.45)
0.520 (0.44 to 0.60)
21 g
150
2
104
44
0
1 (1.00 to 1.00)
0.019 (−0.01 to 0.04)
0.307 (0.23 to 0.38)
ITI
150
78
28
10
34
0.227 (0.10 to 0.35)
0.736 (0.65 to 0.82)
0.587 (0.51 to 0.67)
IPRE (C)
150
76
30
8
36
0.182 (0.07 to 0.30)
0.717 (0.63 to 0.80)
0.560 (0.48 to 0.64)
IPRE (G)
150
94
12
25
19
0.568 (0.42 to 0.71)
0.887 (0.83 to 0.95)
0.793 (0.73 to 0.86)
IPRE
150
101
5
29
15
0.659 (0.52 to 0.80)
0.953 (0.91 to 0.99)
0.867 (0.81 to 0.92)
Here, N defines the total number of samples, TP defines true positive, TN defines true negative, FP defines false positive, FN defines false negative, SE defines sensitivity, SP defines specificity, ACC defines accuracy, and 95% CI defines 95% confidence intervals. Due to space restrictions, we represent the genomic grade index as GGI, 70-gene signature as 70 g, 76-gene signature as 76 g, 112-gene signature as 112 g, invasiveness gene signature as IGS, 21-gene signature as 21 g, and interactome-transcriptome integration as ITI. From the table, the IPRE algorithm achieved superior performance amongst others (as highlighted in bold) in the Desmedt dataset and the Vijver dataset.