Functional Inference of Gene Regulation using Single-Cell Multi-Omics

Abstract

Cells require coordinated control over gene expression when responding to environmental stimuli. Here, we apply scATAC-seq and scRNA-seq in resting and stimulated human blood cells. Collectively, we generate ~91,000 single-cell profiles, allowing us to probe the cis -regulatory landscape of immunological response across cell types, stimuli and time. Advancing tools to integrate multi-omic data, we develop FigR - a framework to computationally pair scATAC-seq with scRNA-seq cells, connect distal cis -regulatory elements to genes, and infer gene regulatory networks (GRNs) to identify candidate TF regulators. Utilizing these paired multi-omic data, we define Domains of Regulatory Chromatin (DORCs) of immune stimulation and find that cells alter chromatin accessibility prior to production of gene expression at time scales of minutes. Further, the construction of the stimulation GRN elucidates TF activity at disease-associated DORCs. Overall, FigR enables the elucidation of regulatory interactions across single-cell data, providing new opportunities to understand the function of cells within tissues.

Decoding gene regulation in the mouse embryo using single-cell multi-omics

“…We find that the in silico knock-out of Brachyury disrupts the transition from NMP to Somitic mesoderm ( Figure 4g ). Although this result was expected based on previous findings (Guibentif et al, 2021), it demonstrates how GRNs inferred from unperturbed single-cell multi-omics data have the potential to provide functional insights into cell fate transitions (Kartha et al, 2021).…”

“…Single-cell multimodal technologies have huge potential for the study of gene regulation (Chen et al, 2019; Clark et al, 2018; Luo et al, 2022; Ma et al, 2020; Zhu et al, 2019, 2021). In particular, the ability to link epigenomic with transcriptomic changes allows the inference of gene regulatory networks (GRNs)(Aibar et al, 2017; Davidson and Erwin, 2006; Kamimoto et al, 2020; Kartha et al, 2021; Materna and Davidson, 2007). GRNs are able to capture the interplay between TFs, cis-regulatory DNA sequences and the expression of target genes (Garcia-Alonso et al, 2019; Levine and Davidson, 2005; Stadhouders et al, 2018), and can hold predictive power of cell fate transitions and gene perturbations (Kamimoto et al, 2020).…”

“…GRNs are able to capture the interplay between TFs, cis-regulatory DNA sequences and the expression of target genes (Garcia-Alonso et al, 2019; Levine and Davidson, 2005; Stadhouders et al, 2018), and can hold predictive power of cell fate transitions and gene perturbations (Kamimoto et al, 2020). Methods that derive GRNs from single-cell genomics data have been developed (Aibar et al, 2017; Fleck et al, 2021; Kamimoto et al, 2020; Kartha et al, 2021) and applied to the developing fly brain (Janssens et al, 2022) but similar analyses of mammalian development are lacking. In addition, GRN inference relies on accurate TF binding data, yet limited knowledge of TF binding exists for early embryonic development due to limitations in experimental methods such as ChIP-seq or CUT&RUN, which require large numbers of cells (Skene and Henikoff, 2017) and faithful antibodies.…”

The continuum of Drosophila embryonic development at single-cell resolution

“…To complement the NNLS, we applied a recently developed tool, FigR ( 37 ), to further facilitate gene regulatory network (GRN) reconstruction. Because multi-omic ATAC-RNA data from the same cell are required for this task, we first integrated our two independent assays for all cells from 10 to 12 hours using canonical correlation analysis (CCA), identifying the most likely ATAC-RNA cell pairs using geodesic distance–based pairing ( 37 ) within the common CCA space.…”

“…To complement the NNLS, we applied a recently developed tool, FigR ( 37 ), to further facilitate gene regulatory network (GRN) reconstruction. Because multi-omic ATAC-RNA data from the same cell are required for this task, we first integrated our two independent assays for all cells from 10 to 12 hours using canonical correlation analysis (CCA), identifying the most likely ATAC-RNA cell pairs using geodesic distance–based pairing ( 37 ) within the common CCA space. Using these pairs as input for GRN inference with FigR, we linked ATAC peaks to their target genes based on peak-to-TSS accessibility correlation and then computed TF motif enrichments for the linked regions, which, together with the TF expression-accessibility correlation, allowed us to define hundreds of putative activators and repressors at this embryonic stage (fig.…”