Statistical Approaches to Analyzing a Large, Multi-Treatment, Time Course Study of Metal Pellets...
Jun 20, 2017 1:21 PM
| Last Modified: Jun 22, 2017 9:52 AM
Statistical Approaches to Analyzing a Large, Multi-Treatment, Time Course Study of Metal Pellets Implanted into Mouse Muscle
Wenjun Bao, Russell D. Wolfinger, and Tzu-Ming Chu, SAS; Edward Perkins, US Army Engineer Research and Development Center; Desmond I. Bannon, Institute of Public Health
In a time course study of the effects of a single implanted pellet (Nickel, Tungsten/Nickel/Cobalt Alloy, Tantalum, or Control) in mice, over 400 Illumina Sentrix MouseRef-8 v1.1 Expression Bead Chip Arrays were generated to assess the effect of treatments on gene expression profiles over time. We present solutions for several data analysis challenges associated with such a large dataset. First, transformation: there are a significant number of genes with negative intensity values. We investigated shifted log and missing value imputation approaches. Second, normalization: there is an obvious batch effect associated with the microarray sample preparation complete date. We explored different normalization methods, including grouped-batch profile (GBP) correction. Third, differentially expressed genes: We used a mixed model ANOVA approach to test their differences.