A new local covariance matrix estimation for the classification of gene expression profiles in high dimensional RNA-Seq data

Recent developments in the next-generation sequencing based on RNA-sequencing (RNA-Seq) allow researchers to measure expression levels of thousands genes for multiple samples simultaneously. In order analyze these kinds data sets, many classification models have been proposed literature. Most existing classifiers assume that are independent; however, this is not a realistic approach real RNA-Seq problems. For reason, some other methods, which incorporates dependence structure between into model, proposed. Quantile transformed Quadratic Discriminant Analysis (qtQDA) recently one those classifiers, estimates covariance matrix by Maximum Likelihood Estimator. However, MLE may reflect genes. we propose new local function estimate be used qtQDA model. This assumes dependencies locally defined rather than complete dependency. The performances classifier two different compared over terms error rates. results show using yields better and increases performance classifier.