Recognition of CRISPR/Cas9 off-target sites through ensemble learning of uneven mismatch distributions

Abstract

Motivation CRISPR/Cas9 is driving a broad range of innovative applications from basic biology to biotechnology and medicine. One of its current issues is the effect of off-target editing that should be critically resolved and should be completely avoided in the ideal use of this system. Results We developed an ensemble learning method to detect the off-target sites of a single guide RNA (sgRNA) from its thousands of genome-wide candidates. Nucleotide mismatches between on-target and off-target sites have been studied recently. We confirm that there exists strong mismatch enrichment and preferences at the 5′-end close regions of the off-target sequences. Comparing with the on-target sites, sequences of no-editing sites can be also characterized by GC composition changes and position-specific mismatch binary features. Under this novel space of features, an ensemble strategy was applied to train a prediction model. The model achieved a mean score 0.99 of Aera Under Receiver Operating Characteristic curve and a mean score 0.45 of Aera Under Precision-Recall curve in cross-validations on big datasets, outperforming state-of-the-art methods in various test scenarios. Our predicted off-target sites also correspond very well to those detected by high-throughput sequencing techniques. Especially, two case studies for selecting sgRNAs to cure hearing loss and retinal degeneration partly prove the effectiveness of our method. Availability and implementation The python and matlab version of source codes for detecting off-target sites of a given sgRNA and the supplementary files are freely available on the web at https://github.com/penn-hui/OfftargetPredict. Supplementary information Supplementary data are available at Bioinformatics online.

CRISPR/Cas Systems in Genome Editing: Methodologies and Tools for sgRNA Design, Off‐Target Evaluation, and Strategies to Mitigate Off‐Target Effects

“…As well as interfering with chromosome stability, off‐target effects may cause loss of functional‐gene activity that causes diverse physiological or signaling abnormalities ( Figure ). It is therefore vital to design an optimum sgRNA for achieving high on‐targeting with no or little possibility for off‐target effects …”

“…This method enhances the efficiency of off‐target site prediction as compared to other computational methods, and can identify more off‐target sites consist with bona fide detections through high‐throughput methods. According to two case studies, it is efficient in the selection of optimal sgRNAs for treatment of certain genetic disorders …”

Induced Methylation in Plants as a Crop Improvement Tool: Progress and Perspectives

“…Off-target binding seems inevitable in large genomes; however, different methods exist to improve CRISPR-Cas9 specificity. Novel machine learning algorithms enable the efficient design of sgRNAs, based on experimentally validated examples [104][105][106]. Limiting the amount of both Cas9 and sgRNA could minimize off-target modifications, as higher concentrations of these increase the risk of binding to sites containing mismatches [107,108].…”

DL-CRISPR: A Deep Learning Method for Off-Target Activity Prediction in CRISPR/Cas9 With Data Augmentation

“…With the definition, 957 unique reliable off-targets and 5,435 ''detected'' off-targets were obtained from in vitro and cellbased assays, from which we deleted sgRNA 'TCATCCTC-CTGACAATCGATAGG' on gene CCR5_9 with its off-target site, as there is only 1 sample for this sgRNA. Then we downloaded off-targets obtained by PCR assays [18] and got 215 more samples. We integrated the reliable off-targets got above without repetition to construct the final positive dataset.…”

“…However, in recent years, the boom of the machine learning breeds a lot of methods in this area, which improves the off-targets prediction accuracy effectively. Abadi et al developed CRISTA to detect the offtargets probability, which uses the Random Forest to learn a regression model [16], Listgarten et al proposed Elevation to predict off-targets activities, which relies on scoring and aggregating scores machine learning models [17], Peng et al presented an SVM ensemble learning method to determine the off-targets propensity of a sgRNA [18], and Chuai et al implemented DeepCRISPR to design optimal sgRNA as well as predicting the off-target profile with deep learning [19].…”