How would you rate the specificity of siRNA/double stranded RNA? What is the chance that a siRNA designed for a specific gene (house keeping gene) will target any other gene having little or no homology?
A RISC loaed with an antisense strand of an siRNa, or a miRNA, has at least two principle modes of targeting mRNA expression. The "siRNA" pathway results in mRNA cleavage by the Ago compound in RISCs, and is typically regarded as involving near-complete complementarity of siRNA and target mRNA. By this mechanism, the 20 or so nucleotides of an siRNa would be perfectly suficent for absolute specificity. However, there is also the so-called "miRNA" mode, which involves a seed region of as little as a 6bp match with the target mRNA, and thus can be sufficient to suppress gene expression. The information provided in 6bp can provide a discrimination by the factor 4x4x4x4x4x4=4096 which is rather low in comparison to the human genome of some 10^9. In practice (i.e. in your cell culture experiments), the specificity is considerably higher, but still siRNA companies devote huge resources to identify off-target effects.
As you can see, I fail to give you a "real" number to work with, because there is no absolute number yet. It will depend on the details of the experiment that you plan. Try some literature....PMID:21652644; PMID:18048158; PMID:17608023