Refined sgRNA efficacy prediction improves large- and small-scale CRISPR–Cas9 applications

Abstract

Genome editing with the CRISPR–Cas9 system has enabled unprecedented efficacy for reverse genetics and gene correction approaches. While off-target effects have been successfully tackled, the effort to eliminate variability in sgRNA efficacies—which affect experimental sensitivity—is in its infancy. To address this issue, studies have analyzed the molecular features of highly active sgRNAs, but independent cross-validation is lacking. Utilizing fluorescent reporter knock-out assays with verification at selected endogenous loci, we experimentally quantified the target efficacies of 430 sgRNAs. Based on this dataset we tested the predictive value of five recently-established prediction algorithms. Our analysis revealed a moderate correlation (r = 0.04 to r = 0.20) between the predicted and measured activity of the sgRNAs, and modest concordance between the different algorithms. We uncovered a strong PAM-distal GC-content-dependent activity, which enabled the exclusion of inactive sgRNAs. By deriving nine additional predictive features we generated a linear model-based discrete system for the efficient selection (r = 0.4) of effective sgRNAs (CRISPRater). We proved our algorithms’ efficacy on small and large external datasets, and provide a versatile combined on- and off-target sgRNA scanning platform. Altogether, our study highlights current issues and efforts in sgRNA efficacy prediction, and provides an easily-applicable discrete system for selecting efficient sgRNAs.

The CRISPR/Cas revolution continues: From efficient gene editing for crop breeding to plant synthetic biology

“…Using Drosophila as an experimental system, Ren et al () showed that a GC content of over 50% is beneficial only within the six nucleobases positioned immediately upstream of the PAM. Then again, Labuhn et al () reported a significant effect of the GC content in the domain comprising positions ‐8 through to −17 upstream of the PAM in human cells.…”

Optimized sgRNA design by deep learning to balance the off-target effects and on-target activity of CRISPR/Cas9

“…These tools are designed to assist researchers in the selection of best target sites by helping them exclude undesirable targets based on predicted low efficiency or a high potential for off-target effects. They could be broadly divided into two groups, on-target cleavage efficiency tools (6,(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38)(39) and off-target activity tools (18,28,33,(40)(41)(42)(43)(44)(45)(46)(47)(48)(49)(50)(51). In the on-target cleavage efficiency evaluation tools, researchers focus on identifying the gRNA sequence features that contribute to target cleavage efficiency.…”

SpCas9 activity prediction by DeepSpCas9, a deep learning-based model with unparalleled generalization performance

“…CRISPR-Cas, a prokaryotic adaptive immune system, has been harnessed for genome editing in various species and cell types, including human cells [1][2][3][4][5][6] . The ability to accurately predict SpCas9 activity is important for applications of genome editing [7][8][9][10][11][12] . So far, several computational models that predict SpCas9 activity have been developed based on data sets of phenotypic changes of gene-edited cells 7,9,[11][12][13][14][15][16][17] or medium-sized data sets of SpCas9-induced indel frequencies obtained by episomal plasmid-based library-on-library approaches 10,18,19 .…”

“…The ability to accurately predict SpCas9 activity is important for applications of genome editing [7][8][9][10][11][12] . So far, several computational models that predict SpCas9 activity have been developed based on data sets of phenotypic changes of gene-edited cells 7,9,[11][12][13][14][15][16][17] or medium-sized data sets of SpCas9-induced indel frequencies obtained by episomal plasmid-based library-on-library approaches 10,18,19 . However, the generalization performances of these models have been limited 20 , possibly because the quality and size of the training data sets were not ideal.…”

“…The majority of models were developed using Cas9 activity data sets generated by phenotypic analysis of gene knockouts 7,9,[11][12][13][14][15][16][17] , which can be biased by the function of the corresponding genes and can include false-negatives in which indels are introduced at the target sequences but do not induce functional knockouts 11 ; furthermore, for three models, the sizes of the SpCas9-induced indel frequency data sets were just mediumsized 10,18,19 .…”