Finding associations in a heterogeneous setting: Statistical test for aberration enrichment

摘要: Abstract Most two-group statistical tests are implicitly looking for a broad pattern such as an overall shift in mean, median or variance between the two groups. Therefore, they operate best settings where effect of interest is uniformly affecting everyone one group versus other. In real-world applications, there many scenarios heterogeneous. For example, drug that works very well on only proportion patients and equivalent to placebo remaining patients, disease associated gene expression dysregulation occurs cases whereas have levels indistinguishable from controls considered gene. these examples with heterogeneous effect, we believe using classical may not be most powerful way detect signal. this paper, developed test targeting effects demonstrated its power controlled simulation setting compared existing methods. We focused problem finding meaningful associations complex genetic diseases omics data expression, miRNA DNA methylation. simulated real data, showed our complementary traditionally used able disease-relevant genes which would detectable previous approaches.