The development of high-throughput biological technologies have enabled researchers to simultaneously perform analysis on thousands of features (e.g., genes, genomic regions, and proteins). The most common goal of analyzing high-throughput data is to contrast two conditions, to identify “interesting’’ features, whose values differ between two conditions. How to contrast the features from two conditions to extract useful information from high-throughput data, and how to ensure the reliability of identified features are two increasingly pressing challenge to statistical and computational science. This dissertation aim to address these two problems regarding analysing high-throughput data from two conditions.