Learning Directed Acyclic Graphs Integrating Bulk Tumor and Single Cell RNAseq Data

• Activity Number: 442 - Methods for Single-Cell and Microbiome Sequencing Data
• Type: Topic Contributed
• Date/Time: Thursday, August 6, 2020 : 10:00 AM to 11:50 AM
• Sponsor: Biometrics Section
• Author(s): Shrabanti Chowdhury* and Jie Peng and Pei Wang
• Companies: Icahn school of Medicine at Mount Sinai and University of California Davis and Icahn Medical School at Mount Sinai
• Keywords: causal association; regulatory networks; directed acyclic graphs; aggregation; prior

Abstract

Inferring gene regulatory network (GRN) based on -omics profiles is commonly pursued in biological studies. Among different approaches of constructing (GRN), directed acyclic graph (DAG) models are commonly used to infer causality of the regulatory relationships among interacting genes in a complex biological system. Recently single-cell RNA-sequencing (scRNA-seq) technologies have made it possible to provide insights into transcriptional networks among cell populations based on the sequencing of a large number of cells. In this work we propose joint learning of multiple cell type-specific DAGs through aggregation in scRNAseq data. Aggregation strategy helps to reduce false positives. In DAG structure learning edge directions are not always identifiable without external information. We utilize methylation/protein data from bulk tumor profiling that provide prior information on regulatory directions to enhance the robustness of the network construction.