Prediction of CRISPR/Cas9 single guide RNA cleavage efficiency and specificity by attention-based convolutional neural networks

Abstract

CRISPR/Cas9 is a preferred genome editing tool and has been widely adapted to ranges of disciplines, from molecular biology to gene therapy. A key prerequisite for the success of CRISPR/Cas9 is its capacity to distinguish between single guide RNAs (sgRNAs) on target and homologous off-target sites. Thus, optimized design of sgRNAs by maximizing their on-target activity and minimizing their potential off-target mutations are crucial concerns for this system. Several deep learning models have been developed for comprehensive understanding of sgRNA cleavage efficacy and specificity. Although the proposed methods yield the performance results by automatically learning a suitable representation from the input data, there is still room for the improvement of accuracy and interpretability. Here, we propose novel interpretable attention-based convolutional neural networks, namely CRISPR-ONT and CRISPR-OFFT, for the prediction of CRISPR/Cas9 sgRNA on- and off-target activities, respectively. Experimental tests on public datasets demonstrate that our models significantly yield satisfactory results in terms of accuracy and interpretability. Our findings contribute to the understanding of how RNA-guide Cas9 nucleases scan the mammalian genome. Data and source codes are available at https://github.com/Peppags/CRISPRont-CRISPRofft .

Effective use of sequence information to predict CRISPR-Cas9 off-target

“…Deep learning models trained on imbalanced data tend to achieve high accuracies for the majority class. However, the learning models generally perform worse for the minority class, which is noteworthy in this case [25] . Data imbalance is a common problem for off-target prediction, and efficient computational techniques can help address the issue [27] .…”

Prediction of activity of CRISPR-Cpf1 guide RNA via in vivo high-throughput screening

“…To further improve the accuracy and efficiency of guide RNA design, we expect that more datasets from well-designed experiments across species could be generated, and more comprehensive algorithms such as machine and deep learning (MDL) methods could be applied. [38][39][40]In addition, improvement of the specificity of guide RNA to reduce offtarget effects is a clinical demand, thus our platform combining high-throughput screening and analysis of computational methods may be used to address this issue.…”

AI applications in functional genomics

“…In particular, in the gRNA design process, it is crucial to optimize the engineered sequence towards specific interaction with the editing target ( on-target activity) while minimizing unintended interactions with other genomic sites ( off-target activity), which may arise from sequence similarity with the genuine target. Various ML methods and DL methods have been developed to optimize gRNA design and predict both on-target and off-target activity, including: CRISTA [170] , an RF-based regression model that scores the propensity of a genomic site to be cleaved by a given gRNA; DeepCRISPR [171] , a computational platform that uses data augmentation technique to expand the training dataset of experimentally validated gRNA sequences and feeds two CNNs (one for on- and one for off-target activity prediction), with gRNA representations produced by pre-trained autoencoders; CROTON [172] , an end-to-end framework based on deep multi-task CNNs and neural architecture search to predicting CRISPR-Cas9 editing outcomes; and the complementary tools CRISPR-ONT and CRISPR-OFFT [173] , attention-based CNNs trained to predict gRNA on- and off-target activities, respectively.…”