Abstract Text: Single-cell RNA sequencing (scRNA-seq) data has revolutionized our understanding of systemic perturbations to organismal physiology, revealing gene expression at the level of individual cell types. However, despite the increased information content and dimensionality of single-cell data, the relevance of genes to a perturbation is still commonly assessed through differential expression analysis. This approach provides only a one-dimensional, high-level perspective of the transcriptomic landscape, risking the oversight of tightly controlled genes characterized by modest changes in expression but with profound downstream effects due to strong connectivity with other genes (e.g., transcription factors or signaling molecules). In this study, we develop a novel in silico quantitative method to identify perturbation-relevant genes. In the context of disease, such genes may be interpreted as possible drivers of disease phenotypes and/or valuable targets for therapeutic intervention. Our approach, Gene Expression Network Importance eXamination (GENIX), is a gaussian graphical-based model that infers cell-type-specific gene association networks to uncover condition-relevant gene programs and target genes through comparative analysis. To demonstrate the effectiveness of GENIX, we utilize a publicly available dataset. In particular, we analyze influenza vaccine-induced immune responses in peripheral blood mononuclear cells (PBMCs) collected longitudinally from recovered COVID-19 male and female patients. This analysis recapitulates and sheds light on the mechanistic underpinnings of gender differences in response to immunization. In summary, our methodology represents a promising avenue for identifying novel targets that could assume pivotal roles in systems biology, thereby expanding the scope of perturbation analysis and target discovery beyond differentially expressed genes.
