Gene expression microarray missing value imputation and its effects in downstream data analysis

DNA microarray is a high throughput gene profiling technology that has been employed in numerous biological and medical studies. These studies require complete and accurate gene expression values which are not always available in practice due to the so-called microarray missing value problem. In this dissertation, most of the existing microarray missing value imputation methods are reviewed and discussed. In these missing value imputation methods, the (normalized) root mean squared error is commonly adopted as a standard measurement of the imputation quality. However, considering that the imputed expression values are for downstream data analyses, we propose to use the microarray sample classification accuracy in addition to (normalized) root mean squared error, to measure the missing value imputation quality. Our extensive comparative study between seven missing value imputation methods circulate our conjecture that the sample classification accuracy is a more appropriate way for measuring the microarray missing value imputation quality.