LAUSR

To the Editor: We recently reported Digenome-seq (digested genome sequencing), a method for in vitro identification of potential off-target sites, and we evaluated the specificity of CRISPR– Cas9 (refs. 1,2) and CRISPR–Cpf1(ref. 3) endonuclease by wholegenome sequencing. Digenome-seq pinpoints the exact location of double-strand break (DSB) sites by recognizing specific patterns of aligned reads. However, the analysis pipeline presented in our previous report required extensive manual interaction and produced several large intermediate files, resulting in a long running time. Here, we present a redesigned analysis tool for Digenome-seq data that runs on web browsers. The core algorithm of the tool is written in C++ and compiled to asm.js (http://asmjs.org/), a preoptimized subset of JavaScript. Users can instantly perform the complete analysis in an ordinary web browser (Supplementary Note 1) with fast execution speed without uploading any data to a server and without local tool installation. In our benchmark, the full analysis for 100 GB of BAM file took 3 h for whole analysis on Intel i5 3570k central processing unit in a single thread. With the new tool, a user submits both the mockand nucleasetreated BAM files (Fig. 1a) together with target sequences and required parameters (Supplementary Note 2). The result page is then displayed and updated in real time. Next, the tool finds patterns of aligned reads that characterize DSB sites. 80-bp flanking sequences around each site are retrieved from the Ensembl database4 and aligned to the target sequence using a semiglobal alignment algorithm for user convenience. Each site is reported along with the alignment, the number of DSB sites per chromosome (Fig. 1b) and in an interactive plot (Fig. 1c). The web tool is freely available at http://www.rgenome.net/digenome-js.