Entering edit mode
10.8 years ago
pld
5.1k
Suppose I have a whole tissue composed of multiple cell types. I homogenize the tissue and use half for an array experiment and the other half I separate using FACs. I would like to demonstrate that each cell type has a unique expression pattern. Rather than comparing every cell type with every other cell type, would it be sound to demonstrate that a given cell differs significantly from the background (the whole tissue expression)?
Two cell types that both differ from the background doesn't necessarily mean they differ from each other. Perhaps a better way to represent the global expression pattern is to do a pair-wise correlation distance for all cell type and then plot them as a tree.
When you say "each cell type", do you mean cell types that you are defining in your FACS analysis? Are you trying to establish that cell populations as defined by FACS have unique expression patterns? Have you considered doing your expression profiling on FACS-sorted populations, as opposed to doing expression profiling on a heterogenous mixture? Deconvolving cellular populations using expression data from a cell mixture is challenging, although there have been a number of publications describing methods for approaching that.
That would indeed be the goal. I am interested in what sorts of models can be developed from mixed cell types either in cell culture or in tissue explants. Originally I was thinking of using the whole tissue expression as a background distribution of expression, then from that same tissue use FACS to sort cells and measure their expression. Then determine if a given FACS cell type is different than the whole tissue background. My thinking was that I could use this to determine which cells are driving which parts of the background (whole tissue) expression. In other words, which cells are driving the phenotype of the tissue you are interested in.
However, I think that it would be much more effective to just look at the FACS tissue. From a pratical standpoint, I'd rather not waste half the sample on an array if it risks driving a given cell population too low to use.