Subtle gene expression differences between individual cells are informative of gene regulation

Single-cell RNA sequencing studies on gene co-expression patterns could yield important regulatory and functional insights, but have so far been limited by confounding factors such as differentiation or cell cycle. We overcome this limitation by applying single-cell RNA sequencing to a population of virtually identical mESCs. This yields meaningful gene co-expression patterns that are informative for gene regulation and function. We find that in mESCs nuclear proximity has the highest effect size while shared transcription factor action impacts most gene pairs. We also show that gene co-expression in this highly homogeneous cell population can be used for regulator and functional prediction. Next we applied a novel technology for RNA and protein co-profiling in single cells (SPARC) to hESCs and find that transcription factor protein but not RNA predicts target expression. We also show that discrepancies between RNA and protein variation can be linked to translation efficiency. Finally, we integrate these data with single-cell small RNA data from mESCs to elucidate previously suggested alternative miRNA functions.