Cell-type specific gene expression profiles are needed for many computational methods operating on bulk RNA-Seq samples, such as deconvolution of cell-type fractions and digital cytometry. However, the gene expression profile of a cell type can vary substantially due to both technical factors and biological differences in cell state and surroundings, reducing the efficacy of such methods. Researchers from the Chalmers University of Technology investigated which factors contribute most to this variation. They evaluated different normalization methods, quantified the variance explained by different factors, evaluated the effect on deconvolution of cell type fractions, and examined the differences between UMI-based single-cell RNA-Seq and bulk RNA-Seq. They investigated a collection of publicly available bulk and single-cell RNA-Seq datasets containing B and T cells, and found that the technical variation across laboratories is substantial, even for genes specifically selected for deconvolution, and this variation has a confounding effect on deconvolution. Tissue of origin is also a substantial factor, highlighting the challenge of using cell type profiles derived from blood with mixtures from other tissues. The researchers also show that much of the differences between UMI-based single-cell and bulk RNA-Seq methods can be explained by the number of read duplicates per mRNA molecule in the single-cell sample. This work shows the importance of either matching or correcting for technical factors when creating cell-type specific gene expression profiles that are to be used together with bulk samples.
Improvement in correlation between 10x Chromium and bulk from regressing out covariates
A. Gene expression for cortex 1 from the EVAL dataset plotted as 10x vs bulk. The orange line represents a perfect correlation. B. Gene expression for cortex 1 from the EVAL dataset after regressing out the differences in UMICF and GC content between 10x and bulk using a loess fit, which improves the correlation. C. Average Pearson correlation coefficient between 10x data and bulk in log scale after regressing out technical covariates (UMI copy fraction, transcript length, GC content and GC content tail), using linear or loess regression. The correlation shown is the average of the correlations from cortex 1 and 2 of the EVAL dataset. The error bars represent the confidence interval, based on bootstrapping of genes to include in the correlation.