Project Abstract In early development and over the lifetime of a human, the genome of every somatic cell will eventually accumulate hundreds of mutations during multiple cell divisions. Although most somatic mutations are predicted to be non-functional, it is known for a long time that some of the somatic mutations, including single nucleotide variants (SNVs), copy number variants (CNVs), translocations, etc., may cause serious diseases like cancer. In the past decade, more and more studies suggested that somatic mutations may also play important roles in milder complex diseases, such as autism. However, although single-cell or ultra-deep whole genome sequencing (WGS) technologies can now identify many rare somatic mutations, these technologies tell little about the consequences or mechanisms of somatic mutations. In fact, unless a somatic mutation causes significant clonal expansion, characterizing the molecular functions of a somatic mutation in its native tissue context is extremely challenging. In general, WGS protocol precludes most of the commonly pursued epigenomic technologies such as ATAC-seq and ChIP-seq. We recent demonstrated that using a novel deep-learning-based pipeline named DeepLoop, we can upgrade the super sparse single cell Hi-C maps to kilobase resolution, which may serve as a robust readout of genome activity. This motivates us to optimize a technology named Dip-C to simultaneously map somatic mutations and 3D genome from single cells. If successful, the project will deliver a long needed multi- OMIC tool for SMaHT network. We will test Dip-C in both model cell line and human tissues and verify its unique capability to resolve how somatic mutations may affect a small number of cells in large population or complex tissue.

Project Narrative The need to understand the functions of rare somatic mutations calls for multi-OMIC technologies that can scan somatic mutations and map epigenome simultaneously. Here we will optimize a method to generate both whole genome sequencing and 3D genome (Hi-C) data from single cells. Using a novel deep-learning-based pipeline, we can upgrade the sparse Hi-C maps to kilobase resolution, which serves as a robust readout for genome activity.