Introduction

In eukaryotic organisms, RNA interference (RNAi) is the sequence-specific gene silencing that is induced by double-stranded RNA (dsRNA) homologous to the silenced gene. In the cytoplasm of mammalian cells, long dsRNAs (>30 nt) can activate the potent interferon and a protein kinase-mediated pathway, which lead to non-sequence-specific effects that can include apoptosis (Kumar and Carmichael, 1998). Elbashir and coworkers (2001a) made the important discovery that small interfering RNAs (siRNAs) of about 21 nt specifically inhibit gene expression, because siRNAs are too short to activate the interferon or protein kinase pathway. The silencing by synthetic siRNAs is transient. This limitation can be overcome by stably expressed short hairpin RNAs (shRNAs), which are processed by Dicer into siRNAs (Paddison et al., 2002; Brummelkamp et al., 2002). However, it was recently reported that shRNA vectors can induce an interferon response (Bridge et al., 2003).

Because target recognition presumably depends on Watson-Crick base pairing, the RNAi machinery is widely believed to be exquisitely specific. As a reverse genetic tool, RNAi has set the standard in high-throughput functional genomics (Barstead, 2001; Tuschl, 2003). RNAi has also become an important tool in the identification and validation of drug targets in preclinical therapeutic development (Thompson, 2002; Appasani, 2003). Furthermore, RNAi-based human therapeutics are under development.

Initial empirical rules have been established by the Tuschl lab for the design of siRNAs. However, large variation in the potency of siRNAs is commonly observed, and often only a small proportion of the tested siRNAs are effective.