2018/09/21 by Harry Gray, Gwenaël G. R. Leday, Gray, Harry +6
Biochemistry, Genetics and Molecular Biology · #Applications (stat.AP) #Bioinformatics and Genomic Networks #FOS: Computer and information sciences #Gene expression and cancer classification #Metabolomics and Mass Spectrometry Studies #Methodology (stat.ME)
paper · pdf · doi:10.48550/arxiv.1809.08024
openalex publication_date 2018/09/21 · openalex created_date 2022/08/03 · openalex updated_date 2026/07/28
Linear shrinkage estimators of a covariance matrix --- defined by a weighted\naverage of the sample covariance matrix and a pre-specified shrinkage target\nmatrix --- are popular when analysing high-throughput molecular data. However,\ntheir performance strongly relies on an appropriate choice of target matrix.\nThis paper introduces a more flexible class of linear shrinkage estimators that\ncan accommodate multiple shrinkage target matrices, directly accounting for the\nuncertainty regarding the target choice. This is done within a conjugate\nBayesian framework, which is computationally efficient. Using both simulated\nand real data, we show that the proposed estimator is less sensitive to target\nmisspecification and can outperform state-of-the-art (nonparametric)\nsingle-target linear shrinkage estimators. Using protein expression data from\nThe Cancer Proteome Atlas we illustrate how multiple sources of prior\ninformation (obtained from more than 30 different cancer types) can be\nincorporated into the proposed multi-target linear shrinkage estimator. In\nparticular, it is shown that the target-specific weights can provide insights\ninto the differences and similarities between cancer types. Software for the\nmethod is freely available as an R-package at http://github.com/HGray384/TAS.\n