Skip to main content

Table 3 Comparison with other algorithms

From: Test on existence of histology subtype-specific prognostic signatures among early stage lung adenocarcinoma and squamous cell carcinoma patients using a Cox-model based filter

Algorithm

# of genes (AC/SCC)

Accuracy

AUC

p-value

A: Comparison with other algorithms on the microarray data itself

Cox-model filter

32(32/12)

74.56(76.38/69.05)

79.82/79.57

2.74 × 10-13

Coxpath separate

142(108/40)

81.66(83.47/76.19)

90.95/87.85

0

glcoxph separate

11(1/10)

66.27(60.63/83.33)

69.99/91.42

3.91 × 10-2

B: Further validation those signatures on the RNA-seq data

Cox-model filter

30(30/11)1

56.8(52.86/61.82)

52.19/68.14

0.305

Coxpath separate

126(97/32)1

52(50/54.56)

47.29/67.36

0.272

glcoxph separate

9(1/8)1

55.20(51.43/60)

57.21/56.31

6.86 × 10-2

  1. Note: In this comparison, we used 1,889 genes with extra barcode-frequency filtering and their actual expression values. 1There are some genes missing from the resultant signatures since the RNA-seq data is on a different platform from microarray data.