I have more than 20 different bioprojects with transcrptomic data under different stress condition. Before performing meta-analysis, I have performed individual differential expression analyses to understand the expression of specific class of protein family(not whole proteome) when plant are treated with some stress condtion to understand the role of that specific protein family. Now, in some bioprojects, I have got NIL DEGs of that specific protein family.
When performing a meta-analysis to combine p-values from individual differential expression analyses, I am confused with the decision to include or exclude bioprojects that show no differentially expressed genes (DEGs), as my primary goals of my analysis is to identifying that specific protein family that are consistently responsive to environmental stress in plants using meta-analysis approach in one specific plant family.
The whole point of meta-analysis is that you get an idea which results are reproducible across many studies. Hence, having no DEGs is fine in individual studies when you can (with some confidence) say that it is not a technical issue.
For example, having no DEGs because there is a number of batch effects that cannot be corrected for, or due to heavy undersequencing or very limited statistical power, then one might want to exclude that study, as nu DEGs represent a technical rather than a biological issue. If no DEGs in a "proper" study is the results though, twhile other studies have some or many DEGs then this means that there is variation in the effect of whatever treatment is done to the specimen, and this indeed should be icluded into the meta-analysis. It will tell in the end whether the treatment effect has a reproducible effect across many datasets.
