Forum:Differential expression analysis for samples from peripheral blood

Question

[Zhilong Jia]

Are there any special methods or trick for differential expression analysis of samples from blood? After dealing with a few GEO datasets where samples are from peripheral blood, I found the adjust p-values are always almost one. Probably, it's due to the heterogeneity. Any suggestion or comments are welcome. Thank you.

[Ahill]

The amount of differential expression you'll see in the blood of course depends on the disease/biology. In diseases where the site of action is far from blood, there may simply be no differential expression to detect in the circulation. Heterogeneity, as you mention, also dilutes expression differences observed in whole blood. In some diseases, like lupus, there are changes that tend to come only from a subset of circulating cell types, but they are large enough that they can be easily seen even in whole blood. Two approaches to deal with heterogeneity have been (a) computational and (b) wet lab cell sorting. Computationally, many authors have recommended de-convolution methods to isolate cell-type specific changes with more sensitivity. You'll see several threads on Biostars about those methods. There have been some successes, but those methods are also not foolproof, and sometimes require information about the cell proportions in your blood samples, information which most GEO peripheral blood data sets will unfortunately not have. In the wet lab, cell sorting to isolate specific cellular populations can be used. A concern with that approach is that the sorting process can also stimulate expression changes that may perturb your measurements.