UC Berkeley

Methods for genome-wide gene expression characterization have proven transformative in the study of biology, facilitating the identification of new types of cells, the unraveling of regulatory and signaling pathways, and the discovery of novel pharmaceutical targets for the treatment of human diseases. Beginning with microarrays and accelerated by the introduction of next-generation sequencing, the technology has progressively increased in its scale, sensitivity, and accuracy. Now, with the recent advent of single-cell RNA-seq (scRNA-seq), it is possible to survey the full heterogeneity of complex tissue. This work deals with the analysis of scRNA-seq data and the development of computational methods to extract new biological insights from single-cell experiments.

Chapter 1 provides some background on the experimental developments enabling single-cell RNA-seq as well as existing methods for the analysis of single-cell data products. Chapter 2 develops an algorithm for the functional interpretation of scRNA-seq data through the use of previous high-throughput experimental results. As a departure from existing methods, this procedure can be utilized with both linear and non-linear dimensionality reduction methods and does not require a pre-partitioning of cells into discrete clusters. In Chapter 3 this approach is further extended for the systematic extraction of informative features (genes) that underlie the heterogeneity of scRNA-seq experiments and the organization of these features into co-regulated gene modules. We additionally demonstrate that this approach can be used with multi-modal datasets to identify genes that vary in accordance with a similarity map specified by a different modality - such as spatial or lineage measurements. Chapter 4 details the development of open source computational tools which make these methods accessible to the larger research community and provide additional capabilities in data exploration, visualization, and collaboration. Finally, in Chapter 5, a case study is described in which single-cell RNA-seq is used to investigate the relevance of the Interleukin 23 receptor in a model of T helper cell type 1 driven autoimmunity.