A Database-Based Two-Phase Algorithm For Efficient and Complete Detection of siRNA Off-Target Homology

Since the discovery of RNA Interference (RNAi), a cellular phenomenon in which a small double stranded RNA induces the degradation of its sequence specific target mRNA, using a computer-aided software tool to help design functional small interfering RNA (siRNA) or small hairpin RNA (shRNA) has become a standard procedure in applying RNAi to silence a target gene. A critical consideration in siRNA design is to avoid any possible off-target effect, i.e. to avoid sequence homology with untargeted genes. Though BLAST is the most powerful sequence alignment tool, it can overlook some significant homologies. Therefore, Smith-Waterman algorithm is the only approach that can guarantee to find all possible mismatch alignments that may cause off-target effect. However, Smith-Waterman alignment suffers from its inefficiency in searching through a large sequence database. A two-phase search algorithm was previously reported in which the first phase is used to identify local regions where the second phase, a bona fide Smith-Waterman alignment, is absolutely needed. Though such a two-phase homology search can improve the efficiency up to two orders of magnitude over the original SmithWaterman alignment algorithm, it is still not efficient enough to be used alone for siRNA off-target homology search over a large sequence database. In this paper, we propose several improvements that dramatically speed up the reported two-phase algorithm while still guaranteeing the complete identification of siRNA off-target homologies.