My lab previously asked me to a meta-analysis to look at a list of genes and see whether they increased or decreased expression between timepoint 1 and timepoint 2. To do this, I took three datasets where labs had collected data at timepoint 1 and timepoint 2, and used rankProd to determine which genes change expression consistently across the three datasets.
Now, my lab is asking for a different thing: they want me to make a bar graph of the expression of the list of genes at each of the timepoints. It would be a straightforward thing to plot if I were using data from only one dataset. What do I do if I want to use three microarray datasets, though? Is there any way to normalize expression across the three labs so that I can take the mean across three datasets for each gene in a single timepoint, and then plot that? Is this a job for ComBat, or is the problem much thornier? If it is thornier, what exactly are the issues with cross-array normalization/does anyone have a citation I can show my lab in case they don't believe me that statistical problems exist?