Evaluation of off-target and on-target scoring algorithms and integration into the guide RNA selection tool CRISPOR

By Maximilian Haeussler, K. Schönig, Hélène Eckert, A. Eschstruth, J. Mianné, J. Renaud, S. Schneider-Maunoury, Alena Shkumatava, L. Teboul, Jim Kent and 2 more. Genome Biology, 2016

1,521 citations recorded by the source corpus.

The success of the CRISPR/Cas9 genome editing technique depends on the choice of the guide RNA sequence, which is facilitated by various websites. Despite the importance and popularity of these algorithms, it is unclear to which extent their predictions are in agreement with actual measurements. We conduct the first independent evaluation of CRISPR/Cas9 predictions. To this end, we collect data from eight SpCas9 off-target studies and compare them with the sites predicted by popular algorithms. We identify problems in one implementation but found that sequence-based off-target predictions are very reliable, identifying most off-targets with mutation rates superior to 0.1 %, while the number of false positives can be largely reduced with a cutoff on the off-target score. We also evaluate on-target efficiency prediction algorithms against available datasets. The correlation between the predictions and the guide activity varied considerably, especially for zebrafish. Together with novel data from our labs, we find that the optimal on-target efficiency prediction model strongly depends on whether the guide RNA is expressed from a U6 promoter or transcribed in vitro. We further demonstrate that the best predictions can significantly reduce the time spent on guide screening. To make these guidelines easily accessible to anyone planning a CRISPR genome editing experiment, we built a new website (http://crispor.org) that predicts off-targets and helps select and clone efficient guide sequences for more than 120 genomes using different Cas9 proteins and the eight efficiency scoring systems evaluated here.

Subfields: Genetics, Molecular Biology

View Paper: https://www.semanticscholar.org/paper/df2ad5d82b08d9b1d9642f7730e0a79453fab7f7

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Genetics

In Molecular Biology

Citations in the corpus

The UCSC Genome Browser database: 2017 update (2016, neutral): “Secondly, the ‘View > In External Tools’ menu provides an option to export the current browser region to external tools (e.g., Ensembl, NCBI MapView (37), CRISPOR (38)).”
The UCSC Genome Browser database: 2018 update (2017, neutral): “The new track is designed to assist in the search for appropriate guide sequences by listing potential binding sites for the CRISPR/Cas9 complex that are near transcribed regions (2).”
Evaluation and Design of Genome-Wide CRISPR/SpCas9 Knockout Screens (2017, background): “Though a variety of small- and medium-scale experiments have been used to guide gRNA selection algorithms (Doench et al. 2016; Haeussler et al. 2016; Heigwer et al. 2014, 2016; Hsu et al. 2013; Liu et al. 2016; Park et al. 2016), to our knowledge our design is based on the largest available set of…”
Delivering CRISPR: a review of the challenges and approaches (2018, result): “However, caution is still needed, as shown in a recent study by Haeussler et al. (2016) that compared predictions from several sgRNA design tools with experimental results published in eight SpCas9 off-target studies.”
CRISPOR: intuitive guide selection for CRISPR/Cas9 genome editing experiments and screens (2018, methodology): “While we were developing the website, many design criteria were published, so we compared them, published the comparison (4) and designed CRISPOR such that it recommends scores that are most accurate for a particular assay.”
DeepCRISPR: optimized CRISPR guide RNA design by deep learning (2018, methodology): “niques like GUIDE-seq [22], Digenome-seq [16, 23, 24], High-throughput genome-wide translocation sequen-”
Rethinking drug design in the artificial intelligence era (2019, background): “For example, genome-wide off-target screening has much lower sensitivity than analysis of a pre-defined off-target area 25,26 .”