Adaptively Weighted Meta-analysis in -omics Applications

Abstract

In this talk, I will present an adaptively weighted Fisher's method that was developed to account for gene-specific heterogeneous differential expression (DE) signal across studies in transcriptomic meta-analysis. We will show that the method is asymptotically Bahadur optimal and the adaptive weights provide improved interpretation and further biological investigation in omics applications. A fast algorithm has been developed for accurate p-value calculation that is useful for multiple comparison of thousands of genes (or millions of SNPs). Bootstrap technique is used to assess a confidence score for the weights. Simulations and applications of prostate cancer, lung cancer and mouse metabolism data will be presented.