A Nonparametric Statistical Method for More Reproducible Biomarker Detection

• Activity Number: 155 - Research Reproducibility for Precision Medicine: From Controlled Experiments to Real-World Evidence
• Type: Topic Contributed
• Date/Time: Monday, July 29, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Wei Zhuang* and Luísa Camacho and Camila Silva and Huixiao Hong
• Companies: NCTR/U.S. FDA and NCTR/U.S. FDA and NCTR/U.S. FDA and NCTR/U.S. FDA
• Keywords: Research Reproducibility ; Precision Medicine ; Nonparametric Statistical Method ; Biomarker Detection; Gene Expression ; qPCR

Abstract

Introduction: Recent studies have revealed that microRNAs are promising biomarkers for detection of diseases. The technique of qPCR (Quantitative Real-time Polymerase Chain Reaction) is often used to measure the expression levels of genes and microRNAs. Besides complete data, investigators have observed technically inevitable incomplete qPCR data. Investigators usually set incomplete observations equal to the maximum number of qPCR cycles (MC), apply the complete-observation method (CO), or choose not to analyze targets with incomplete observations (CNA). The three methods tend to cause biased inference and decrease research reproducibility. Methods: We propose a nonparametric statistical cycle-to-threshold method (CTOT), which incorporates qPCR-specific features and is built around extracting information from all subjects, censored or not. Results and Conclusions: Our simulations and application show that CTOT may improve the power of detecting differential effects of a treatment over the existing methods without excess type I errors. CTOT helps leverage qPCR technology, increase the power to detect novel biomarkers, and improve research reproducibility for precision medicine.