If you’re interested in questions about effective science funding, single cell transcriptomics is an interesting field to study. Quick primer for those unfamil

If you’re interested in questions about effective science funding, single cell transcriptomics is an interesting field to study. Quick primer for those unfamiliar with single cell transcriptomics (ELI18 level): it’s a technique where you take one cell at a time and measure which genes are on (expressed) and measure the level of expression. The resulting data is _very_ high dimensional. Typical experiments these days result in millions of cells and tens of thousands of gene expression values (after much processing). Since single cell RNA-seq (scRNA-seq) emerged there’s been a focus to “map” the transcriptome of every possible cell type, in a quest to understand more about different cell types and the relationships between them. Huge amounts of grant money has poured in from large science funders, notably including the Chan Zuckerberg Initiative. The technology has improved rapidly over the last decade, and it’s now common to see scRNA-seq experiments tacked onto papers/studies that wouldn’t have such data 3+ years ago. There is a negative side to the rise of scRNA-seq data however. Since the data is inherently high-dimensional, it can be difficult to interpret. Every single cell transcriptomic experiment produces results, regardless of whether there is interesting underlying biology to be revealed. A common analysis flow in the field is to apply dimensionality reduction (PCA, tSNE, UMAP) to the data and then interpret the resulting clusters/mapping. This is really problematic as it’s trivial to show that random data can produce clusters. I don’t want to paint the picture that all groups analyse scRNA-seq incorrectly. However, the rise in availability of the technology far outpaced the field’s ability to analyse the data. These issues don’t apply to the Human Cell Atlas project, which has specialists analysing the data and much more internal/external scrutiny on the data. However, there’s still a question of lacklustre scientific output. Large projects such as the Human Cell Atlas are perfect for advancing scientists careers. They are too big to fail. Owing to the difficulty in interpreting scRNA-seq and the mapping/atlas goal of the project, there is no way for it to fail. Any output data is taken as gospel. I don’t want to be too cynical. Mapping all human cell types is a genuinely interesting scientific question. On the other hand, I’m not sure that the way these mega-projects are structured produces the best science. On the optimistic side, I think mega projects do significantly help technology development. The Human Cell Atlas created huge demand for scRNA-seq technologies and products, much in the same way that the Human Genome Project did before it. I think in the long term that will be the lasting positive impact, rather than the results themselves. Context: I have a PhD in single cell transcriptomics from a lab that went on to get a large grant as part of the Human Cell Atlas project.