Determining gene specificity from multivariate single-cell RNA sequencing data

Multivariate single-cell experiments can assay gene specificity across sex, tissue, age, genotype, and cell state. Existing specificity measures largely overlook multivariate structure or fail to decompose coherently across hierarchical partitions. Ember supplies three entropy-based metrics and is the only method shown to satisfy four desired specificity properties, revealing context-dependent markers, housekeeping genes, strain-driven cell-type switching, tissue-dominated mouse variation, and cell-type-dominated human PBMC variation.