Table 2. Precision of outcome prediction using PAM analysis of exon- or gene-level-derived signatures.
Prediction of OS
|
||||||
---|---|---|---|---|---|---|
Gene–based
|
Exon–based
|
|||||
Prediction accuracy (%) | Sensitivity (%) | Specificity (%) | Prediction accuracy (%) | Sensitivity (%) | Specificity (%) | |
PAM | 83.1 | 84.7 | 82.6 | 83.6 | 86.5 | 82.8 |
PAM59 | 86.1 | 79.6 | 89.4 | 84.9 | 80.8 | 86.9 |
PAM-MDC | 89.2 | 85.2 | 91.4 | 88.8 | 85.2 | 90.7 |
Abbreviations: MDC=maximum divergent course; OS=overall survival; PAM=prediction analyses for microarray.
Values are presented as the fraction of the total for accuracy, sensitivity and specificity. The signature was derived from the PAM feature selection process described in the Materials and methods sections. A published gene set consisting of 59 genes was used to predict patient outcome in PAM59 using either the gene- or exon-based expression data (Vermeulen et al, 2009a). Only patients with MDCs were included in the PAM-MDC analysis to test the signature derived from PAM.