Adjusting confounders in ranking biomarkers: a model-based ROC approach

Abstract

High-throughput studies have been extensively conducted in the research of complex human diseases. As a representative example, consider gene-expression studies where thousands of genes are profiled at the same time. An important objective of such studies is to rank the diagnostic accuracy of biomarkers (e.g. gene expressions) for predicting outcome variables while properly adjusting for confounding effects from low-dimensional clinical risk factors and environmental exposures. Existing approaches are often fully based on parametric or semi-parametric models and target evaluating estimation significance as opposed to diagnostic accuracy. Receiver operating characteristic (ROC) approaches can be employed to tackle this problem. However, existing ROC ranking methods focus on biomarkers only and ignore effects of confounders. In this article, we propose a model-based approach which ranks the diagnostic accuracy of biomarkers using ROC measures with a proper adjustment of confounding effects. To this end, three different methods for constructing the underlying regression models are investigated. Simulation study shows that the proposed methods can accurately identify biomarkers with additional diagnostic power beyond confounders. Analysis of two cancer gene-expression studies demonstrates that adjusting for confounders can lead to substantially different rankings of genes.

Copula-based Partial Correlation Screening: a Joint and Robust Approach

“…Van't Veer et al (2002) was the first to study the breast cancer study involving 97 lymph node-negative breast cancer patients 55 years old or younger, of which 46 developed distant metastases within 5 years (metastatic outcome coded as 1) and 51 remained metastases free for at least 5 years (metastatic outcome coded as 0). This expression data set with clinical variables has been well analysed in many papers for classification (Boulesteix et al (2008), Yu et al (2012), among others).…”

A model‐based multithreshold method for subgroup identification

“…Yu et al performed a screening analysis on this breast cancer data using the receiver operating characteristic–based approach by adjusting for the clinical risk factors. Their methods produced 10 important genes and 10 genes by practical ranking.…”

Multi-threshold accelerated failure time model

“…We re-visit this data set in this section and consider an AFT model between the failure time and 5 most significant genetic markers selected in Yu et al (2012). According to Yu et al (2012), genes 357, 2345, 6267, 6271, and 3653 in the original sample are the most important markers for the survival risk prediction when clinical information is adjusted. We thus set x 1,i = 1 as the intercept, chose the five gene expressions as the regressors x 2,i , .…”