Skip to main content

Table 2 Prognostic gene lists rely on genes correlated with the first (cell cycle) principal component variable1.

From: Cell cycle correlated genes dictate the prognostic power of breast cancer gene lists

   Intercept adjusted PC1 adjusted PC2 adjusted
Data set Gene list Good/Poor2 HR (p-value) Good/Poor2 HR (p-value) Good/Poor2 HR (p-value)
NKI2        
  70 Gene 91/174 5.8 (< 0.0001) 30/235 1.2 (0.67) 90/175 5.7 (< 0.0001)
  Wang 76-gene (ER+) 85/60 4.5 (< 0.0001) 44/101 0.9 (0.84) 89/56 2.7 (0.005)
  Wang 76-gene (ER-) 6/29 0.9 (0.88) 5/30 1.2 (0.78) 8/27 0.5 (0.25)
  Wound Signature 74/221 4.1 (0.0002) 27/268 1 (0.93) 73/22 3.9 (0.0002)
  Sotiriou Grade 92/203 4.7 (< 0.0001) 25/270 0.8 (0.8) 91/204 4.7 (< 0.0001)
  Naderi 131/104 2.5 (0.0006) 123/112 0.4 (0.003)2 128/107 2.4 (0.0014)
KJX64/KJ125        
  70 Gene 38/111 2.8 (0.03) 15/134 0.7 (0.4) 39/110 2.9 (0.03)
  Wang 76-gene (ER+) 53/36 1.6 (0.32) 35/54 1.2 (0.7) 61/28 1.8 (0.24)
  Wang 76-gene (ER-)4 14/5 n/a 16/3 n/a 12/7 n/a
  Sotiriou Grade 46/143 2.9 (0.03) 35/154 1.7 (0.24) 46/143 2.7 (0.03)
  Naderi 71/56 3.9 (0.004) 69/58 1.6 (0.24) 72/55 3.9 (0.004)
Wang        
  70 Gene 49/197 2.6 (0.005) 22/224 0.9 (0.41) 44/202 2.3 (0.015)
  Wang 76-gene (ER+) 45/84 6.3 (0.0005) 30/99 2 (0.11) 49/80 5.6 (0.0003)
  Wang 76-gene (ER-) 26/16 11.7 (0.0001) 29/13 8.2 (0.0001) 15/27 4.9 (0.03)
  Sotiriou Grade 57/229 2.5 (0.007) 21/265 0.7 (0.25) 49/237 2 (< 0.0001)
  Naderi 106/110 2.4 (0.0008) 105/111 0.9 (0.66) 105/111 3.1 (< 0.0001)
  1. 1. Hazard ratios (HR) represent the change in risk for tumors classified as having a poor prognosis versus tumors classified as having a good prognosis using a univariate proportional hazards analysis. Gene expression data were independently adjusted using simple linear regression analysis for either an intercept only, the principal component representing the first correlated gene cluster identified (PC1) or the second correlated gene cluster identified (PC2) in each data set. Univariate HRs that are significantly (p < 0.05) greater than 1 are shown in bold.
  2. 2. Values in the Good/Poor column indicate the number of tumors assigned to the good and poor prognosis groups, respectively. The total number of tumors per gene list will vary depending upon the number of tumors used in training sets.
  3. 3. The fact that this HR is significantly less than 1 indicates that the classifier did not appropriately classify tumors into appropriate good and poor prognosis groups.
  4. 4. There were too few events among good and poor prognosis groups to perform statistical analyses in the KJX64/KJ125 data set.