Cas9-chromatin binding information enables more accurate CRISPR off-target prediction

Abstract

The CRISPR system has become a powerful biological tool with a wide range of applications. However, improving targeting specificity and accurately predicting potential off-targets remains a significant goal. Here, we introduce a web-based CR: ISPR/Cas9 O: ff-target P: rediction and I: dentification T: ool (CROP-IT) that performs improved off-target binding and cleavage site predictions. Unlike existing prediction programs that solely use DNA sequence information; CROP-IT integrates whole genome level biological information from existing Cas9 binding and cleavage data sets. Utilizing whole-genome chromatin state information from 125 human cell types further enhances its computational prediction power. Comparative analyses on experimentally validated datasets show that CROP-IT outperforms existing computational algorithms in predicting both Cas9 binding as well as cleavage sites. With a user-friendly web-interface, CROP-IT outputs scored and ranked list of potential off-targets that enables improved guide RNA design and more accurate prediction of Cas9 binding or cleavage sites.

Figure 1.

(A) Schematics of CROP-IT algorithm. Comparative analysis of prediction performance of CROP-IT (B) Crispr.mit.edu (16); (C) CRISPRSeek (28); (D) E-CRISP (31) and (E) CasOFFinder (29) for two different sgRNAs: Nanog-sgRNA and PhC1-sgRNA. Y-axis indicates the overlap between the maximum number of computationally predicted Cas9 binding sites for each tool and ChIP-Seq identified sites. The number of sites mentioned under the X-axis is top N sites for CROP-IT to match the number of predicted sites output by the other tool. For CasOFFinder, top 100 000 predicted sites (out of ∼6 000 000) were compared with those of CROP-IT (F) Comparison of CROP-IT and Crispr.mit.edu based on overlapping Cas9 binding sites with top 10, 100 and 200 predicted sites for Phc1 and EMX1 sgRNAs.

Figure 2.

Comparative analysis of prediction performances of CROP-IT and CasOT (30) tools. Both CROP-IT and CasOT output substantial numbers of predicted sites. Thus, variable numbers of top predicted sites for each tool (X-axis) are analyzed for overlap with (A) Nanog-sgRNA and (B) PhC1-sgRNA mediated Cas9 bound ChIP-Seq sites from Wu et al. (22). Y-axis indicates the overlap with ChIP-Seq identified sites.

Figure 3.

Incorporating the contribution of chromatin structure to CROPT-IT algorithm. (A) Percent overlap between Cas9 ChIP-Seq sites (n = 2600) with DNase I-seq identified DNase I hypersensitive sites (HS) from 125 different human cell types. Error bars indicate the s.d. of 1000 computational simulations for randomly selected 2600 genomic sites. X-axis display different bins of HS sites according the frequency of observation in different number of cell types. Comparison of outputs from implementing CROP-IT with DNase I chromatin state information and CROP-IT without chromatin information by overlapping (B) Cas9 binding sites and (C) Cas9 cleavage sites with predicted sites. Top 2000, 4000, 7000 and 10 000 computationally predicted sites are analyzed.

Figure 4.

Comparative analysis of prediction performances of CROP-IT with (A) Crispr.mit.edu (16) and (B) E-CRISP (31) on GUIDE-Seq identified cleavage (24). Since each tool outputs different number of predictions, the same of sites for each tool were compared to the same number of CROP-IT predicted sites picked from the top of ranked list. Y-axis indicates the overlap with GUIDE-Seq identified Cas9 cleavage sites. (C) Comparative analysis of prediction performances of CROP-IT and Crispr.mit.edu (16) on cleavage sites identified through experimental HTGTS (25) and Di-genome-Seq (26) approaches for VEGFA site. Y-axis indicates the overlap with sites identified by each study. (D) Comparison of experimentally verified and different number of top sites predicted by CROP-IT. Y-axis indicates the percent overlap with total GUIDE-Seq identified Cas9 cleavage sites (indicated in X-axis). (E) Similar analysis presented in (D) was performed for HTGTS (25) and Di-genome Sequencing (26) identified Cas9 cleavage sites for VEGFA targeted guiding RNA. Y-axis indicates the percent overlap with total experimentally identified Cas9 cleavage sites and CROP-IT predicted top off-target sites (indicated in X-axis).