Skip to main content
. 2013 Feb 1;14:74. doi: 10.1186/1471-2164-14-74

Table 3.

Performance of signature gene sets: cross prediction matrix showing prediction errors

Class comparison
Signature gene set
This study
Berry et al. [[4]]
Maertzdorf et al. [[6]]
  Study 116 gene set Minimal TB-LTBI set Minimal TB-LTBI-HC set Robust 10 gene q-PCR set 86 gene set 393 gene set 11 gene set 5 gene set
TB vs. LTBI
Berry et al.[4]
20.1
19.4
nda
nd
12.5
11.1
13.0
16.3
Maertzdorf et al.[5]
26.5
22.5
nd
nd
19.4
24.5
22.5
21.4
Maertzdorf et al.[6]
19.1
16.9
nd
nd
11.3
10.1
10.2
10.1
This study
11.1
11.1
nd
nd
50.0
50.0
33.0
50.0
average prediction error
19.2
17.5
nd
nd
23.3
23.9
19.7
24.5
TB vs. LTBI vs. HC
Berry et al.[4]
27.6
nd
23.9
34.9
20.1
14.1
27.3
30.4
Maertzdorf et al.[5]
48.6
nd
50.3
52.5
47.4
48.6
50.0
41.8
Maertzdorf et al.[6]
25.6
nd
25.6
48.9
17.8
21.3
26.2
32.8
This study
11.1
nd
14.8
14.8
66.7
74.1
74.1
70.4
    28.2
nd
28.7
37.8
38.0
39.5
44.4
43.8
TB vs. LTBI vs. HC vs. other disease Berry et al.[4] 37.4 nd nd nd 27.6 18.7 28.5 37.2

Columns represent selected gene biomarker sets in the literature sets as well as in our dataset. Rows represent the studies on which the gene biomarker sets displayed in the columns were tested.

a = not determined.