Single-Cell Transcriptomics Unveils Gene Regulatory Network Plasticity

Abstract

Single-cell RNA sequencing (scRNA-seq) plays a pivotal role in our understanding of cellular heterogeneity. Current analytical workflows are driven by categorizing principles that consider cells as individual entities and classify them into complex taxonomies. We have devised a conceptually different computational framework based on a holistic view, where single-cell datasets are used to infer global, large-scale regulatory networks. We developed correlation metrics that are specifically tailored to single-cell data, and then generated, validated and interpreted single-cell-derived regulatory networks from organs and perturbed systems, such as diabetes and Alzheimer's disease. Using advanced tools from graph theory, we computed an unbiased quantification of a gene's biological relevance, and accurately pinpointed key players in organ function and drivers of diseases. Our approach detected multiple latent regulatory changes that are invisible to single-cell workflows based on clustering or differential expression analysis. In summary, we have established the feasibility and value of regulatory network analysis using scRNA-seq datasets, which significantly broadens the biological insights that can be obtained with this leading technology.

Single cell transcriptomic analysis revealed long-lasting adverse effects of prenatal tamoxifen administration on neurogenesis in prenatal and adult brains

“…We have shown that prenatal TAM administration has broad effects from NPC proliferation, and cell differentiation to neural circuit formation. To reveal underlying molecular mechanisms, we carried out gene regulatory network analysis using bigSCale2 toolkit to identify hub genes that TAM employed to manipulate corticogenesis (Iacono et al, 2019). We chose to analyze NPC1/2 and CH clusters because the tight association between VZ NPC proliferation and cortical neurogenesis.…”

“…A total of 3,035 cells (CTL, 1,908 cells; TAM, 1,127 cells) were analyzed to generate a gene expression correlation network using bigSCale2 algorithm (Iacono et al, 2019). We discovered four major network modules in CTL and three modules in TAM treated cells ( Fig.…”

“…The NPC1 (pallial neural progenitors), NPC2 (subpallial neural progenitors), and CH (cortical hem) clusters of each condition (CTL, TAM) were merged and served as the input for inferring the putative gene-to-gene correlated network using bigSCale2 algorithm (Iacono et al, 2019). Specifically, two matrices with 17071 genes' expression counts for 1908 CTL and 1127 TAM cells were used to infer the network by the default parameters (i.e.…”

A bivariate zero-inflated negative binomial model and its applications to biomedical settings

“…Statistical inferences at both individual gene level (Iacono et al (2019) and Yu (2018)) and gene set level, e.g., pathways, can be misleading without considering technical zeros.…”

“…Inference of gene-gene dependence, e.g., the correlation-based method, has been widely used in pathway analysis of bulk RNA-seq data (Zhang and Horvath, 2005), and recently also used in scRNAseq data analysis (Iacono et al, 2019;Yu, 2018;Pont et al, 2019;Van Dijk et al, 2018;Eraslan et al, 2019). However, correlation between two genes with technical zeros in the scRNAseq will not reflect the true gene-gene dependence.…”