Honors Thesis Presentation: "Modeling and Applications of Anchor-Based scRNA-seq Data Integration"

Abstract: Single-cell RNA-seq (scRNA-seq) has advanced our ability to understand cellular functions and biological mechanisms at a high resolution. To conduct more comprehensive and in-depth analyses, large-scale datasets are needed. However, since data are generated by different labs, with different technologies, and under different conditions, various noises may hinder the integration and understanding of the data. Here, I discuss several current computational methods, especially anchor-based methods, designed to tackle this data integration challenge. After analyzing the mechanisms of one such method, Seurat v3, I simulate scRNA-seq data with a location-shift model and identify the effect of large location shifts on the performance of integration. Lastly, I demonstrate an exciting application of anchor-based integration, which enables cross-species comparisons among early embryos. With this application, I show that micropatterned hESCs resemble the early-to-late stage of gastrulation and may contain cells transcriptomically similar to PGCs and amnion cells. Overall, my goal is to illustrate both caveats and prospects of anchor-based single-cell integration strategies.

Host: Blake Thornton

(Access the Zoom Presentation HERE)