Data-driven detection of subtype-specific differentially expressed genes

Abstract Among multiple subtypes of tissue or cell, subtype-specific differentially-expressed genes (SDEGs) are defined as being most-upregulated in only one subtype but not any other. Detecting SDEGs plays a critical role the molecular characterization and deconvolution multicellular complex tissues. Classic differential analysis assumes null hypothesis whose test statistic is subtype-specific, thus can produce high false positive rate and/or lower detection power. Here we first introduce One-Versus-Everyone Fold Change (OVE-FC) for detecting SDEGs. We then propose scaled (OVE-sFC) assessing statistical significance that applies mixture distribution model tailored permutation test. The OVE-FC/sFC was validated on both type 1 error power using extensive simulation data sets generated from real gene expression profiles purified samples. applied to two benchmark samples detected many known previously unknown Subsequent supervised results synthesized bulk data, obtained independent by test, showed superior performance accuracy when compared with popular peer methods.