Dear all, Many thanks for all your response. I have to agree with Huber and Crispin on normalizing the data before combining it. However Laurent, thank you for the interesting codes. After combining the data, I will look at the correlation of the redundant genes (on Unigene terms) for each A-B pair. Thus we can reject the A-B pair whose correlation is 'poor'. 2000+ genes will certainly be a good representation of the A-B agreement. Thank you. -- Adaikalavan Ramasamy