Benchmarking the Analysation and Visualization of Single Cell RNA Sequencing Data

Abstract

This study analyzes and visualizes single cell RNA sequencing (sc-RNA-seq) data to understand how cells of different cell types relate to each other. Adult GTEx single cell RNA data was downloaded from GTEx Portal. Dimensionality reduction techniques like PCA, t-SNE and UMAP are used to visualize the data in 2-D. Cells from eight tissues are grouped by gene expressions and colored by their associated cell types. A composite data visualization shows a UMAP reference map in the middle, surrounded by a cell typology; different levels of cell types and clustering analysis on cells can be examined in the four corners to provide a comprehensive and interpretable view of all the information. As part of the study, we compare different methods of dimensionality reduction on GTEx data and compare embeddings performance for k-means clustering algorithm and a graph based clustering algorithm called Leiden. Since both UMAP and t-SNE create a low-dimensional graph of the data to visualize the data in 2-D/3-D, we can perform network analysis on the final graph to identify outliers and detect communities. We plan to expand the study to cover public HuBMAP single cell RNA Sequencing data and by conducting a user study with subject matter experts. Finally, we will create an automated workflow that can be used by biologists to check the quality of their datasets and gain meaningful information about their data through visualization and advanced analytics.