Calibrated Identification of Feature Dependencies in Single-cell Multiomics

Data-driven identification of functional relationships between cellular properties is an exciting promise of single-cell genomics, especially given the increasing prevalence of assays for multiomic and spatial transcriptomic analysis. Major challenges include dealing with technical factors that might introduce or obscure dependencies between measurements, handling complex generative processes that require nonlinear modeling, and correctly assessing the statistical significance of discoveries. VI-VS (Variational Inference for Variable Selection) is a comprehensive framework designed to strike a balance between robustness and interpretability. VI-VS employs nonlinear generative models to identify conditionally dependent features, all while maintaining control over false discovery rates. These conditional dependencies are more stringent and more likely to represent genuine causal relationships. VI-VS is openly available at , offering a no-compromise solution for identifying relevant feature relationships in multiomic data, advancing our understanding of molecular biology. ### Competing Interest Statement N.Y. is an advisor and/or has equity in Cellarity, Celsius Therapeutics, and Rheos Medicine.