Modeling CRISPR-Cas13d on-target and off-target effects using machine learning approaches

Abstract

A major challenge in the application of the CRISPR-Cas13d system is to accurately predict its guide-dependent on-target and off-target effect. Here, we perform CRISPR-Cas13d proliferation screens and design a deep learning model, named DeepCas13, to predict the on-target activity from guide sequences and secondary structures. DeepCas13 outperforms existing methods to predict the efficiency of guides targeting both protein-coding and non-coding RNAs. Guides targeting non-essential genes display off-target viability effects, which are closely related to their on-target efficiencies. Choosing proper negative control guides during normalization mitigates the associated false positives in proliferation screens. We apply DeepCas13 to the guides targeting lncRNAs, and identify lncRNAs that affect cell viability and proliferation in multiple cell lines. The higher prediction accuracy of DeepCas13 over existing methods is extensively confirmed via a secondary CRISPR-Cas13d screen and quantitative RT-PCR experiments. DeepCas13 is freely accessible via http://deepcas13.weililab.org.

Cas13d-mediated isoform-specific RNA knockdown with a unified computational and experimental toolbox

“…We hypothesize that this will hold true for future RNA-targeting CRISPR systems. Practically, this means that one can use existing prediction tools for junction-targeting gRNA design such as TIGER or other models (Wei et al 2021; Cheng et al 2023).…”

“…To obtain Cheng et al predictions for all junction-spanning gRNAs, we re-trained the DeepCas13 model using LFCs for the 5,726 gRNAs provided in their paper (Cheng et al 2023)(https://bitbucket.org/weililab/deepcas13/src/master/). We then used this model to predict gRNA efficiency for all 2.2 million GENCODE junction-spanning gRNAs.…”

Deep screening of proximal and distal splicing-regulatory elements in a native sequence context

“…The limitation of the one-vector system was also recently noted in an independent study 44 . To overcome this issue, we used two separate lentiviral vectors for dCas13d and gRNA expression, which were sequentially transduced into cells, similar to recent Cas13d-based knockdown screens published during the course of this study [42][43][44][45][46][47][48] . Another key issue that made dCas13d-based screening more challenging, compared to Cas13d-based knockdown screening, is likely the requirement for more stable binding of dCas13d/gRNA and target sites to effectively compete with cognate RNA-binding effectors.…”

“…Targeted cleavage of specific RNA transcripts by the Cas13 protein is achieved by using a guide RNA (gRNA) that consists of a direct repeat (DR) and a 22-30 nucleotide (nt) spacer that is complementary to the target RNA sequence. The Cas13 system has been used to perform high-throughput knockdown screenings of genes that influence cellular fitness, including by targeting circular RNAs and long noncoding RNAs [42][43][44][45][46][47][48] , through the use of a pooled lentiviral gRNA library. In addition, the catalytically dead Cas13 (dCas13) has been used as a programmable RNA-binding module, fused to various effector domains for applications in RNA editing 49 , m 6 A modifications 50 , modulation of splicing 37,51,52 and polyadenylation 53 , live cell RNA imaging 54 , and mapping of protein-RNA interactions through proximity ligation 55 .…”

Advanced approaches of the use of circRNAs as a replacement for cancer therapy

“…On the other hand, the CRISPR/Cas system is another developing and powerful strategy can be used to reduce off-target effects. For instance, using a machine learning methodology, Cheng et al discovered that DeepCas13 effectively differentiates efficient sgRNAs from inefficient ones, and Cas13d′s effect is linked to the guide's on-target cleavage effect in hepatocellular carcinoma [ 264 ]. Moreover, the CRISPR-dCas9 and CRISPR-Cas9 systems can also be used to demonstrate the function of this circular gRNA in vitro.…”