I am trying to integrate multiple different studies into a single meta analysis by combining their raw data to study differential gene expression for the conditions but limiting myself to a single platform namely Affymetrix Human Genome U133 Plus 2.0 Array. I followed the advice given here 1, normalized data with SCAN.UPC and removed batch effect using Combat from sva package. I am getting negative values in the expression data and feel uncomfortable using it. Is there a better method to approach this problem? Also are there any other published studies with similar approach, searching I found most of them had multiple platforms focus? Even what keyword to focus on would be helpful.