Abstract The treatment of cancer suffers from diverse drug responsiveness mainly due to high heterogeneity among cell types/states. Sophisticated gene regulatory mechanisms define and maintain cancer types, and in turn, underlie different cancer evolution trends. Within a global gene regulatory system, a regulon represents a maximal group of genes co-regulated by the same transcription regulator (TR, including transcription factors and long non-coding RNAs). Empowered by single-cell RNA-Seq, substantial gene expression analyses have been carried out to reveal regulon activities and construct gene regulatory networks in melanoma, mammary tumors, etc. A clear and profound understanding of these regulons provides an effective way to characterize the heterogeneous regulatory mechanisms, as it allows us to pinpoint crucial regulators and their targeting genes encoded in diverse cell states along with the development of cancer. Hence, for the first time, we define an alternative regulon (AR) as a group of chromatin accessible genes that co-regulated by the same TR in a cell type. Not surprisingly, a TR can regulate multiple heterogeneous ARs with diverse gene availability and activity in different cell types, making the computational identification of ARs a tremendous challengeable problem. Here, we build a first-of-its-kind web server, CeRIS, to predict cell-type-specific ARs based on single-cell RNA-Seq data. Five steps are included: (i) pre-processing, including gene imputation (DrImpute), normalization (scran), and quality control (Seurat3.0); (ii) representing gene regulatory signals by the in-house left-truncated mixture Gaussian model; (iii) clustering cell types using Seurat3.0 and identifying active gene modules in each cell type based on the in-house biclustering method (QUBIC2); (iv) finding conserved cis-regulatory motif patterns for active gene modules using the in-house tool DMINDA2.0 and predicting ARs using TR-motif databases; and (v) ARs with component genes being uniquely accessible or highly expressed in one specific cell type than other cell types are recognized as cell-type-specific ARs. The performance of CeRIS was evaluated using >30 datasets and it outperformed existing tools in terms of the percentage of CTSRs in predicted regulons, marker gene coverage, and pathway enrichment. An applicative study on multiple myeloma suggested that CeRIS can identify novel onco-CTSRs including oncogenes and driver transcription factors in multiple myeloma. Another integrative study of single-cell multi-omics data from human lung indicated the alternative regulons across cell types that can be used to generate heterogeneous landscapes for gene regulatory mechanisms. We believe such novel characteristics of a regulon can be critically useful for cell reprogramming as it can show the transition mechanisms between cell types influenced by the specific dynamic changes of gene regulations. Citation Format: Qin Ma, Anjun Ma, Cankun Wang, Yang Li, Chi Zhang. Towards cell-type-specific gene regulation in heterogeneous cancer cells [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 4409.
Read full abstract