Benchmark of bioinformatics tools for cell type classification in single-cell RNA sequencing data

In recent years, the advent of single-cell RNA sequencing has revolutionized the transcriptomics research field providing a deeper understanding of cells’ diversity and dynamics in complex tissues and organisms. One of the most critical steps in scRNAseq data analysis workflow is the annotation of cell types, which involves the categorization of cells within datasets into distinct classes, based on some known phenotypes or molecular signatures. The accurate definition of cells’ identities is a complex task: if conducted manually, it can be extremely laborious, time-consuming and prone to errors since it relies on partially subjective decision-making. Moreover, the growing high-dimensionality of data with thousands of cells per experiment, poses new challenges in developing specialized bioinformatics tools. For these reasons, many automated supervised or semi-supervised classification approaches have been adapted to ensure reproducible and consistent results. However, despite the diversity in their basic functioning algorithms, there is a lack of extensive and comprehensive comparisons of such methods, which further complicates the selection of the most appropriate tool. The ultimate goal of this project is to provide a benchmark of some of the existing approaches for cell labeling, in order to systematically compare and evaluate their overall performance across publicly available scRNAseq datasets.