| A useful biomarker should be precise, reliable and specific in distinguishing normal and disease samples. Often, ROC (Receiver Operating Characteristic) algorithms are used to evaluate biomarkers through calculation of sensitivity and specificity. Multiplexed biomarker assays pose a statistical challenge because classical ROC curve analysis is typically based on single markers and pair-marker computation is infeasible. Information from more than one marker could potentially improve the accuracy of existing single marker tests. We describe a novel algorithm that permits the efficient computation of ROC curves for marker-pairs. Further, we demonstrate the application of the algorithm to a clinical data set for the identification of useful marker-pairs. |
|
