The paper introduces ELLA (subcellular expression localization analysis), a novel statistical framework designed to model and detect spatial variation of gene expression at the subcellular level within cells using high-resolution spatial transcriptomics data. ELLA utilizes an over-dispersed nonhomogeneous Poisson process and a unified cellular coordinate system to analyze diverse cellular morphologies and transcriptomics technologies, outperforming existing methods like SPRAWL and Bento in simulations. The application of ELLA across multiple real-world datasets—including mouse liver, mouse embryo, fibroblast, and mouse brain data—consistently reveals that nuclear-enriched genes are often longer and associated with long noncoding RNAs and transcription factors, while cytoplasmic- or membrane-enriched genes frequently encode ribosomal proteins or contain signal peptides. Overall, ELLA is presented as a powerful, robust, and scalable tool for subcellular spatial expression analysis, uncovering new biological insights into mRNA localization and its association with cellular function and disease.

References:

• Wang J X, Zhou X. ELLA: modeling subcellular spatial variation of gene expression within cells in high-resolution spatial transcriptomics[J]. Nature Communications, 2025, 16(1): 9920.