An Adaptive Method for Incorporating New Genomic Markers into an Existing Prognostic Classifier

• Activity Number: 6
• Type: Invited
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: WNAR
• Author(s): Sean Devlin *+ and Irina Ostrovnaya and Mithat Gönen
• Companies: Memorial Sloan Kettering Cancer Center and Memorial Sloan Kettering Cancer Center and Memorial Sloan Kettering Cancer Center
• Keywords: Recursive partitioning ; Prognostication ; Genomic markers

Abstract

Prognostic classification systems help guide clinical care and patient counseling in many malignancies. They are validated and well-integrated into patient care but usually have modest prognostic ability; hence further refinement and improvement of these systems using new genomic panels is an area of active clinical research. In this talk, we propose an algorithm to refine an existing classification system using new genomic markers. The new methodology recursively searches for improvements in the predictive accuracy of a classifier and utilizes cross-validation for tuning parameter selection. Simulation studies evaluate the performance of the algorithm as a function of the number and prevalence of the genomic markers. The proposed algorithm is illustrated in acute myeloid leukemia, where new markers may help refine the existing cytogenetic prognostic system.