Enzymatic amplification is not the solution to profiling of very small cell samples...
The fewer cells that forms the sample you want to profile for gene expression, the more problems you will have. This seemingly obvious statement is as a general rule not taken seriously by the scientific community, it seems. Stochastic variation will be a very prominent feature for any gene, save a few percent of the most highly expressed. And these are generally not what one is looking for. We have recently performed a quantitative study of these effect in BMC Genomics. The study is based on the "TransCount" ANVOA model, jointly developed by scientists at the Norwegian Radium Hospital and Norwegian Computing Center. The model is available as a publication in Nucleic Acids Research. This method holds the promise of obtaining absolute transcript numbers from any single- or dual-channel microarray experiment.