The need to reassess single-cell RNA sequencing datasets: more is not always better

Abstract

Background: The advent of single-cell RNA sequencing (scRNAseq) and additional single-cell omics technologies have provided scientists with unprecedented tools to explore biology at cellular resolution. However, reaching an appropriate number of good quality reads per cell and reasonable numbers of cells within each of the populations of interest are key to infer conclusions from otherwise limited analyses. For these reasons, scRNAseq studies are constantly increasing the number of cells analysed and the granularity of the resultant transcriptomics analyses. Methods: We aimed to identify previously described fibroblast subpopulations in healthy adult human skin by using the largest dataset published to date (528,253 sequenced cells) and an unsupervised population-matching algorithm. Results: Our reanalysis of this landmark resource demonstrates that a substantial proportion of cell transcriptomic signatures may be biased by cellular stress and response to hypoxic conditions. Conclusions: We postulate that the ”more is better” approach, currently prevalent in the scientific community, might undermine the extent of the analysis, possibly due to long computational processing times inherent to large datasets.

The need to reassess single-cell RNA sequencing datasets: the importance of biological sample processing

“…Reassessment of the main cell populations in a large skin dataset reveals the presence of clusters with stress-and hypoxia-related gene signatures By using an unsupervised population-matching algorithm (details in processed notebooks available online 20 ) we observed that in each of the healthy donors analysed by Reynolds et al, 13 at least two independent fibroblast clusters expressed signature markers of the A1, A2, B1 and B2 populations. One set of cells corresponded to the Fb2 population, and the second set corresponded to the Fb1 and Fb3 populations.…”

“…Processed notebooks and AnnData files can be found at: https://doi.org/10.5281/zenodo.4596374. 20 License: Creative Commons Attribution 4.0 International.…”

Challenges and Opportunities for the Translation of Single-Cell RNA Sequencing Technologies to Dermatology

“…In this case, fibroblast subpopulations were characterized by the expression of signature genes COL1A1 and COL1A2, CXCL12, and CCL19, respectively [40]. Of note, a preprint posting a reanalysis of this study postulates that their fibroblast clustering is unreliable due to technical reasons [33].…”

“…Of all solid tissues, the study of skin has gained traction in the single-cell community because it is accessible and read-ily available. However, skin disaggregation-as with any other solid tissue-presents some technical challenges, and this will impact cell viability, as well as the representativeness of the obtained sample [32,33]. To date, most of the single-cell studies on healthy skin are focused either on determining the heterogeneity of fibroblasts from the dermis or understanding transcriptional changes underlying the differentiation of keratinocytes in the interfollicular epidermis.…”

“…Correct analysis of these datasets requires biological expertise and an expert data analyst to understand dataset biases and take the correct decision on how to proceed. Incorrect analyses of datasets-especially large datasets requiring computationally intensive analyses-can result in incorrectly annotated populations or oversimplifications of complex cell states [33]. Moreover, translation of data analysis requires robust and standardized pipelines.…”