Abstract
A gene is defined as essential when its functional loss compromises an organism's viability. Identifying essential genes is critical for identifying the components that regulate a biological system. Advances in gene editing techniques like CRISPR-Cas9 provide a capacity to interrogate a genome to elucidate the genes that are essential. However, these techniques are often applied for a single-cell line and rarely probed at a level of a tissue or organ. The recent availability of large-scale single-cell RNA-sequencing (scRNA-seq) atlases provides an unprecedented opportunity to investigate essential gene expression in a more comprehensive context. Our study leverages information from benchmarking datasets, single-cell tissue atlases, and databases of essential genes, to develop a method, scEssentials, that uses a statistical framework to investigate the robustness and specificity of essential genes across multiple cell types. Using scEssentials, mouse and human models showed consistently high expression and exhibited limited variability across more than 60 cell types. We demonstrate a substantial number of significantly correlated gene pairs that produce densely connected co-expression networks with functional annotation. Finally, we develop a score to quantify the relative essentiality of genes within scEssentials, further validating their significant association with gene mutation frequency and chromatin accessibility. Using ageing as an application, we demonstrate how scEssentials identifies robust gene expression profiles. Only one-fifth of scEssentials genes showed significant ageing-related differential expression among age groups. Collectively, the robustness of scEssentials serves as a reference for analysing scRNA-seq data and provides insight into the heterogeneous nature such as ageing.