Bridging the gap: joint modeling of single-cell 1D and 3D genomics

Advances in single-cell technologies have enabled three-dimensional (3D) genome structure profiling and simultaneous capture of the transcriptome and epigenome within a cell. Quantitative tools are, however, still unable to fully leverage the unprecedented resolution of single-cell high-throughput chromatin

conformation (scHi-C) data and integrate it with other single-cell modalities. To address this challenge, I propose to (1) Develop a single-cell gene-body associating domain (scGAD) scoring system to explore single-cell 3D genomics data in units of genes. (2) Construct machine learning-based models to impute

histone modification and 3D chromatin interaction for simultaneously profiling of each cell's epigenomic features and 3D chromatin architectures. Subsequently, I will develop an epigenomic regulatory score

(ERS) model to infer the cell-type-specific promoter-enhancer regulation programs at the highest singlecell and single-gene resolution. (3) Validate and extend scGAD and ERS pipeline to CAR-T immunotherapy study to gain insights into the impact of distal gene regulation variations on patient responses. In Aim 1, preliminary analysis on human and mouse brain tissues demonstrated that scGAD

extracts gene features agreeing well with the scRNA-seq data from the same system. As a result, scGAD facilitates the projection of cells from 3D genomics data onto reference panels constructed by scRNA-seq embeddings with known cell-type annotations. Hence, scGAD provides an unprecedentedly accessible

and accurate cell type annotation method based on 3D chromatin architectures. Furthermore, the successful integration of cells from different modalities into the same network facilitates information sharing across 3D chromatin structures, the transcriptome, and the epigenome. Aim 2 leverages such multi-modal

networks to build an ERS model. ERS jointly models the histone profiles at the promoter and distal neighborhoods of the target gene and the 3D spatial proximity between them. Therefore, the ERS scores quantify the regulatory effects of distal elements on a per gene and cell basis. Aim 3 will extend the

integration framework in Aim 1 and 2 using scRNA-seq as a multi-modality bridge to CITE-seq data for a deeper annotation, especially for the Peripheral Blood Mononuclear Cells. This enables the in-depth investigation of the apheresis samples from the Acute Lymphoma Leukemia patients to gain insight into

the roles of distal regulatory elements on gene expression and their impact on the CAR-T cell response.