Issues Regarding Biomarker Combination Within ROC Framework: Strategies, Target Functions, and Large Number of Weak Markers

• Activity Number: 278 - Combining Markers for Classification in Practical Tasks
• Type: Invited
• Date/Time: Tuesday, July 31, 2018 : 8:30 AM to 10:20 AM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Lili Tian* and Li Yan and Jingjing Yin and Le Kang
• Companies: SUNY at Buffalo and Roswell Park Cancer Institute and Georgia Southern University and Virginia Commonwealth University
• Keywords: Step-wise method; pair-wise method; Receiver Operating Characteristic curve; Area under ROC curve; Youden index; genomic study

Abstract

In practice, usually multiple biomarkers are measured on the same subject for disease diagnosis. Combining biomarkers into a single score could improve diagnostic accuracy. However, methods to combine multiple biomarkers are either distribution-dependent or computationally clumsy. This problem becomes more challenging when the number of observations is not order of magnitude greater than the number of variables, especially when the involved biomarkers are relatively weak. Furthermore, combination methods usually use AUC (or PAUC) as objective function, however AUC (or PAUC) can have serious drawbacks. This talk will present some recent developments in this field addressing several key issues in biomarker combination within ROC framework.