Consequences and opportunities arising due to sparser single-cell RNA-seq datasets

Abstract

With the number of cells measured in single-cell RNA sequencing (scRNA-seq) datasets increasing exponentially and concurrent increased sparsity due to more zero counts being measured for many genes, we demonstrate here that downstream analyses on binary-based gene expression give similar results as count-based analyses. Moreover, a binary representation scales up to ~ 50-fold more cells that can be analyzed using the same computational resources. We also highlight the possibilities provided by binarized scRNA-seq data. Development of specialized tools for bit-aware implementations of downstream analytical tasks will enable a more fine-grained resolution of biological heterogeneity.

Fig. 1

More cells, more zeros. Binarized scRNA-seq datasets were generated by binarizing the raw count matrix, where zero remains zero and every non-zero value is assigned a one. A Association between year of publication, total number of cells. Scatterplot of the number of cells (log scale) against the date of publication. B Scatterplot of the detection rate (y-axis) against the number of cells (log scale, x-axis). C On the x-axis the Pearson’s correlation coefficient (p) of every cell from the PaulHSC dataset between the binarized and normalized expressions. On the y-axis the product of the detection rate and the variance of the non-zero values (q). α is the Pearson’s correlation coefficient between these values p and q across all cells. D Boxplots of the α-values for all 56 datasets grouped by technology. One dataset (LawlorPancreasData) was excluded as α-value (α = 0.42) for this dataset was a clear outlier

Fig. 2

A, B Cells plotted against the first two principle components of the AD dataset [20]. A PCA based on binary representation, and B PCA based on count representation. UMAP generated from data presented with C the binary-based PCs and D the count-based PCs. Colors indicate annotated cell type. E, H UMAP based on the count based PCs, in which cells are colored according to the binary representation of the marker genes AQP4 (E) and TNR (H) which are known markers for astrocytes and OPCs respectively [21]. F, G Similar as E and H but showing the normalized expression of the marker gene. I The performance (median F1-score) of cell type identification by SingleR [22] and scPred [23] when applied to binary (binarized data), normalized (normalized expression), and shuffled (shuffled normalized expression) for 22 datasets