To the Editor: Programmable nucleases such as Cas9 RNA-guided engineered nucleases (RGENs) 1 enable gene knockout in cultured cells and organisms by producing site-specific DNA double-strand breaks, whose repair via error-prone nonhomologous end joining gives rise to small insertions and deletions (indels) at target sites, often causing frameshift mutations in a protein-coding sequence2. The efficiency of this method can be reduced by in-frame mutations via microhomology-mediated end joining3, 4 (Fig. 1a). Here we present a computer program that assists in the choice of Cas9 nuclease, zinc-finger nuclease and transcription activator–like effector nuclease (TALEN) target sites, using microhomology prediction to achieve efficient gene disruption in cell lines and whole organisms. First we examined the mutations induced by ten TALENs and ten RGENs in human cells via deep sequencing (Supplementary Table 1 and Supplementary Methods). We focused our analysis on deletions because they are much more prevalent than insertions (98.7% vs. 1.3%, respectively, for TALENs; 75.1% vs. 24.9% for RGENs) and because microhomology is irrelevant for insertions. In aggregate, microhomologies of 2–8 bases were found in 44.3% and 52.7% of all deletions induced by TALENs and RGENs, respectively (Supplementary Fig. 1 and Supplementary Table 2). Thus, 43.7%(0.987× 0.443) and 39.6%(0.751× 0.527) of all the mutations induced by TALENs and RGENs, respectively, were associated with microhomology. At a given nuclease target site, the effect of these microhomologyassociated deletions can be predicted. In the extreme cases,(i) all deletions cause frameshifts in a protein-coding gene or (ii) no deletions cause frameshifts. In contrast, one-third of microhomologyindependent deletions result in in-frame mutations. Assuming that~ 60% of indels are microhomology independent on average, the fraction of in-frame mutations at a given site can range from 20%(60%/3+ 0%) to 60%(60%/3+ 40%), a threefold difference between the two extreme cases. Because most eukaryotic cells are diploid rather than haploid, the fraction of null cells carrying two outof-frame mutations can range from 16%(0.40× 0.40) to 64%(0.80× 0.80), depending on the choice of target site. A careful analysis of indel sequences also revealed that the frequency of microhomology-associated deletions depends on both the size of the microhomology and the length of the deletion (SupplementaryFig. 2). On the basis of these observations, we developed a simple formula and a computer program (Supplementary Fig. 3) to predict the deletion patterns at a given nuclease target site that are associated with microhomology of at least two bases (Fig. 1band Supplementary