Pitfalls in analysis and interpretation of single-cell RNA-seq data in cancer

Abstract Single-cell and single-nuclei RNA-seq (sc/snRNA-seq) have become a central approach in cancer research, and their widespread impact has been enabled by various computational tools developed specifically for sc/snRNA-seq analysis. Nevertheless, proper analysis and interpretation of sc/snRNA-seq data requires significant expertise, and the inadequate utility of certain computational methods may lead to dubious results. To mitigate these issues, it is important to recognize the limitations of sc/snRNA-seq data, the assumptions of common methods, and to perform robust analysis. Here, I describe common pitfalls in cancer sc/snRNA-seq analysis and discuss ways to overcome them. Among others, this includes a discussion of potential errors in statistical analysis, in inference of chromosomal aberrations, in trajectory analysis, and in signature-based analysis of bulk RNA-seq data. This review may help readers to avoid common pitfalls and to perform informative analysis and careful interpretation of sc/snRNA-seq datasets in cancer.