Online Program Home
My Program

Abstract Details

Activity Number: 638 - Bayesian Methods for Time-To-Event and Frailty
Type: Contributed
Date/Time: Thursday, August 2, 2018 : 8:30 AM to 10:20 AM
Sponsor: Section on Bayesian Statistical Science
Abstract #328605
Title: Adjusting for Handling Effects in Microarray Data for Prognostic Biomarker Discovery and Survival Risk Prediction
Author(s): Ai Ni* and Mengling Liu and Li-Xuan Qin
Companies: Memorial Sloan Kettering Cancer Center and New York University and Memorial Sloan Kettering Cancer Center
Keywords: Handling Effects; Microarray Data Analysis; Prognostic Biomarker Discovery; Regularized Cox Proportional Hazards Regression; Survival Risk Prediction

Microarray has been a highly useful tool for identifying prognostic genes to be incorporated into the prediction of survival risk based on patients' gene expression profile. A well-known caveat of microarray data measurements, however, is that they are often contaminated with handling effects arising from non-uniform experimental handling. It has been shown in the context of group comparison and classification that the negative impact of handling effects can be prevented effectively by careful study design and moderately by post-hoc data normalization. Research is still lacking on how handling effects impact biomarker discovery and risk prediction for survival outcomes. In this study, we conducted extensive simulations to elucidate this impact under both univariate analysis and multivariate regularized Cox proportional hazards regression. We propose several strategies to reduce the impact of handling effects and numerically evaluate their performance. Finally, we apply the proposed strategies to a microRNA study on ovarian cancer from Memorial Sloan Kettering Cancer Center to demonstrate their performance in biomarker discovery and risk prediction for progression-free survival.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program