A0781
Title: Unsupervised learning approaches for bulk and single cell genomics
Authors: Sushmita Roy - University of Wisconsin-Madison (United States) [presenting]
Abstract: Advances in genomic technologies have substantially expanded the repertoire of high-dimensional datasets that measure different types of modalities such as the transcriptome, epigenome and three-dimensional genome organization. As these datasets are sparse in addition to being high-dimensional, an open challenge is to effectively analyze these datasets to extract meaningful low-dimensional patterns that reflect interpretable cell, gene or region clusters. Non-negative Matrix Factorization (NMF) is a popular dimensionality reduction approach that has been used for diverse types of biological and non-biological datasets. In this talk, I will present extensions of NMF to tackle two problems in regulatory genomics. First, extensions of NMF for understanding the three-dimensional organization of the genome and its role in phenotypic variation. The results show that NMF is a powerful approach for analyzing 3D genome organization from Hi-C assays that can recover biologically meaningful topological units and also enable us to smooth sparse Hi-C datasets and also identify dynamics in 3D genome organization. In the second part of my talk, I will present applications of NMF and its extensions to analyze single-cell RNA-seq datasets. We will present an application of NMF for deriving robust cell clusters from scRNA-seq data of the developing hindbrain and spinal cord and how NMF can be extended to handle multi-sample data to identify common and context-specific cell clusters.
