Project Summary Gene co-expression network analysis is a key inference tool for detecting latent relationships invisible to standard workflows of clustering and differential expression analysis. Such a network approach was instrumental in bulk RNA-seq analysis to link genes with biological processes. Despite the remarkable progress in method development for scRNA-seq analysis, there are no established best practices for constructing robust gene co-expression networks from scRNA-seq data. With the wide availability of scRNA-seq technology, population-scale scRNA-seq datasets across multiple subjects and time points/perturbations are emerging. Although the immediate analyses of these datasets focus on the standard analysis of clustering and differential expression, leveraging the power of scRNA-seq at the co-expression network level has the potential to unlock genes converging into key disrupted regulatory pathways. Network-level variation, when associated with phenotypic variation (e.g., severity of response to virus), can reveal critical biological insights. Such an advancement presents constructing personalized dynamic co-expression networks and identifying dynamic gene modules by taking into account the individualized nature of the networks as the next critical challenge in population-scale scRNA-seq analysis. This proposal will address these challenges in two aims. Aim 1 will develop a de-biasing approach to estimate gene-gene correlations from scRNA-seq data with safeguards against low sequencing depth, data sparsity, and varying numbers of cells and detect correlations that are otherwise obscured by technical limitations. Aim 2 will innovate a regularized spectral clustering method that takes in as input co-expression networks of genes at the subject and time/perturbation levels and infers dynamic gene modules. Both aims will be accomplished through a combination of methodological development, theoretical analysis, data-driven simulation, computational analysis, and experimental validation. Successful completion of the project will deliver foundational methods and software that are applicable to a wide range of scRNA-seq datasets and are uniquely positioned for analyzing population-scale scRNA-seq data.

Project Narrative Emerging population-scale scRNA-seq datasets pose unique challenges for constructing and analyzing co-expression networks because of the variation in the numbers of cells per subject, overall sequencing depths of the cells, and other technical artifacts. This project seeks to develop biologically motivated, statistically rigorous, and computationally feasible approaches to address these challenges inherent to scRNA- seq-based co-expression networks. Resulting rigorously benchmarked and validated methods will enable associating personalized co-expression network features to phenotypic variance.