A Computational Pipeline to Control the Quality and Reduce Contamination in Single Retinal Ganglion Cells

Abstract

Single-cell transcriptome profiling has transformed our understanding of cellular heterogeneity. However, single-cell data with poor quality can impede proper identification of distinct cell populations and subsequent biological interpretations. In this study, we present a customized computational approach to control the quality and reduce contaminations in single-cell transcriptome profiling of retinal ganglion cells (RGCs). We leverage domain knowledge and statistical methods to effectively eliminate various sources of contaminants for identification of RGC types and subtypes. We show that our end-to-end computational pipeline improves the accuracy and reliability of single-cell transcriptome profiling of RGCs and enhances the biological interpretations. To show the effectiveness of our pipeline, we use 5,994 RGCs captured from retinas of mouse using Fluidigm technology as a benchmark dataset and compare with widely used quality control tools. Further, we introduce seven candidate F-RGC subtype markers that we identified after applying our introduced pipeline on the benchmark dataset. Our customized quality control pipeline could enable retinal single RGC probing with more granularity, leading to new insights into RGC-related visual diseases and development of therapeutic approaches.