Unleash the potential of single-cell genomic analyses using publicly available bulk data

Abstract: Single-cell genomic technologies are rapidly transforming biomedical research. However, the power of these technologies is often constrained by the sparsity and high noise level of their data. Moreover, while single-cell multi-omic technologies evolve fast, currently most single-cell datasets only contain one data modality. In this talk, I will discuss how publicly available bulk genomic data can be used to help unleash the potential of single-cell analyses. First, I will show that by using the rich functional genomic data in ENCODE to train prediction models, one can predict global cis-regulatory landscape from single-cell RNA-seq data when single-cell regulome (e.g. single-cell ATAC-seq) data are not available. A big data regression approach BIRD was developed for this high-dimensional prediction task. Second, I will briefly discuss how publicly available bulk chromatin accessibility data can improve signal reconstruction of single-cell ATAC-seq data. Third, I will show that the ENCODE bulk human and mouse functional genomic data can be used to improve cross-species integration of single-cell data. These examples demonstrate the value of publicly available bulk data in single-cell analyses and highlight opportunities for data scientists to make contributions to this fast-evolving field.