Figure 5From: A program to identify prognostic and predictive gene signaturesIdentification of a predictive gene signature using PIMS. A) The optimal signature length was identified by using ROC analysis for each n-length signature (n = 1-200, n = 82). B) ROC analysis of 82 gene signature evaluated using LOOCV (AUC: 0.73, p < 0.000001). C) LOOCV of the 82 gene signature stratified patients into responder and non-responder groups (OR: 4.0, p < 0.00001). D) ROC analysis of 82 gene signature in validation cohort (AUC: 0.70, p = 0.0012). E) 82 gene signature stratified validation cohort patients into responder and non-responder groups (OR: 4.4, p = 0.002).Back to article page