Limits to the inference of gene regulation from bulk tissue expression data

Motivation Thousands of studies have used co-expression analysis of bulk tissue samples to probe gene regulation. However, the extent that intracellular regulatory signals are present in these data is unclear. Specifically, we lack clarity of the factors that promote or impede the propagation of intracellular regulatory signals from the single cell level to the bulk tissue level. To bring these issues into focus, we developed a novel computational simulator, grounded in real data, to explore the theoretical relationship between events in single cells and bulk tissue expression profiles, and clarify the conditions required for the propagation of intracellular regulatory signals in complex tissues such as the brain. Results Our simulator first generates single cell expression profiles and subsequently samples and aggregates these single cells to produce bulk tissue expression profiles. Using this framework, we found that there are very specific and unlikely conditions under which intracellular dynamic regulatory signals can be propagated to the bulk tissue level. For the most part, such regulatory relationships, however strong at the single cell level, are unlikely to be detectable. Our results provide a quantitative explanation for why regulatory network inference from co-expression has proved challenging - even with the assistance of other data modalities - and gives the scientific community a set of tools to further explore these issues in both single-cell and bulk tissue data. Availability and implementation All relevant data are within the manuscript and supplementary files. The code for all data analyses and generation of figures are available on GitHub ( https://github.com/PavlidisLab/coex-simulation ). A copy of the data has been deposited in Borealis, the Canadian Dataverse Repository (https://borealisdata.ca/dataset.xhtml?persistentId=doi:10.5683/SP3/2CWXY6).