Statistically principled feature selection for single cell transcriptomics.

Background: The high dimensionality of data in single cell transcriptomics (scRNAseq) requires investigators to choose subsets of genes ("feature selection") for downstream analysis (e.g., unsupervised cell clustering). The evaluation of different approaches to feature selection is hampered by the fact that, as we show here, the difficulty of feature selection can vary greatly, depending on the dataset being analyzed.

Results: For routine cell type identification, even randomly chosen features can perform well, but for cell type differences that are subtle, both number of features and selection strategy matter strongly. We present a simple feature selection method grounded in an analytical model that allows for interpretable delineation of how many and which features to choose, facilitating identification of biologically meaningful rare cell types. We compare this method to default methods in scanpy and Seurat, as well as SCTransform, showing how greater accuracy can often be achieved with surprisingly few, well-chosen features.

Conclusions: Feature selection is a critical step in scRNAseq for downstream analyses. We explore the pitfalls that can arise from incautious feature selection and present a statistical method to facilitate improved outcomes.