Skip to main content
Premium Trial:

Request an Annual Quote

Bioinformatics Method Could Improve Detection of Off-Target CRISPR Activity


CHICAGO – Israeli geneticists and bioinformaticists have developed a novel computing method that they say is superior to earlier tools in detecting off-target activity in CRISPR-Cas9 genome editing. This technology, called CRISPector, can spot potential off-target modifications and prevent translocations, increasing confidence in gene editing, according to the researchers.

The scientists, led by Ayal Hendel, a CRISPR-Cas9 researcher at Bar-Ilan University, and Zohar Yakhini, a computer scientist at the Interdisciplinary Center of the Arazi School of Computer Science at Technion, the Israel Institute of Technology, described their software in a newly published paper in Nature Communications.

According to the paper, the software tool can identify off-target events at a rate as low as 0.1 percent, giving it the potential to significantly improve the accuracy of genome-editing measurements and to enable the broader use of genome editing in biotechnology and therapeutic applications, the authors noted.

Hendel said that the shortcomings of CRISPR are widely known in the genetics world.

One barrier is lack of accuracy, which can be hard to detect because errors in the next-generation sequencing process itself often occur near double-stranded breaks, creating noise that hampers the ability of bioinformatics software to tell the difference between NGS errors and genome edits. "You need to identify the signal beyond the noise," Hendel said.

Another potential pitfall is cutting the genome in the wrong place, causing double-stranded breaks that lead to translocation, which, in turn, could cause diseases including cancer and infertility, the authors noted. Hendel said that current assays for identifying translocations can be unwieldy.

"We wanted to actually take the current methodologies that people are using right now to identify off-target activity and to use the same data to also be able to identify translocations, which people couldn't find before," he explained.

In the Nature Communications paper, the researchers said that current algorithms that analyze off-target activity fail to quantify this activity, are not all that adept at separating signal from noise (particularly where there are low editing rates), and cannot detect translocations. "CRISPector facilitates the statistical analysis of NGS data from multiplex-PCR comparative experiments to detect and quantify adverse translocation events," they wrote.

"We applied statistical models that came by viewing the reads from the treatment and seeing the reads in the controlled experiments where there are no edits," Hendel said. "By applying these statistical methods, you pretty much can now distinguish between the noise and the signal in the treatment experiment, which is really important."

CRISPector processes indel and translocation genome-editing data from an assay performed on paired treatment and control sets using a combination of multiplex PCR and NGS to detect, evaluate, and quantify CRISPR edits. "As such, CRISPector addresses the validation, quantification, and characterization of off-target activity in genomic sites pre-identified by unbiased discovery approaches such as GUIDE-seq," the researchers wrote.

The tool does not read the entire genome, but rather targets areas in the genome that are "prone to off-target activity," according to Hendel.

This happens in the confines of "mock" controlled experiments that do not involve CRISPR editing. The researchers then compared these mock experiments to actual gene editing.

For this study, the researchers evaluated CRISPector against 226 off-target sites. They showed translocations in treated RAG1, RAG2, and EMX1 loci in a HEK293-Cas9 cell line, then validated translocation events with Droplet Digital PCR and had three lines of validation for indels.

Hendel described CRISPR-Cas9 as "DNA scissors." The CRISPR component navigates the scissors to the right location, while the Cas9 part actually makes the cut. Quality-control tools such as CRISPector evaluate the accuracy of the CRISPR guide.

"We don't need to go design a new CRISPR technology which is going to be more accurate," Hendel said. "I can replace the guide RNA with a better guide RNA … that gives you the same desired, on-target editing activity, but now with lower off-target activity."

The researchers said that there are at least 10 previously developed bioinformatics tools for estimating gene editing activity rates.

Most CRISPR quality-control tools Hendel has seen are from academic settings, and even the handful of commercial ones have roots in academia. For example, Cardea Bio, which is affiliated with Keck Graduate Institute in Claremont, California, is a codeveloper of CRISPR-BIND, a rapid and highly sensitive tool to characterize guide RNA and CRISPR-Cas interactions.

The Nature Communications paper listed several existing approaches for identifying potential off-target sites, including Genome-wide, Unbiased Identification of DSBs Enabled by Sequencing (GUIDE-seq), Circularization In Vitro Reporting of Cleavage Effects by Sequencing (CIRCLE-seq), Selective Enrichment and Identification of Adapter-Tagged DNA Ends by Sequencing (SITE-seq), and Discovery of In Situ Cas Off-Targets and Verification by Sequencing (DISCOVER-seq).

Another tool, rhAmpSeq, is a commercial product from Integrated DNA Technologies, or IDT, introduced in 2019 that can validate and quantify potential off-target sites, the authors noted. Four of the listed authors of the new Nature Communications paper represented IDT.

A team at Boston Children's Hospital developed a method called High-Throughput, Genome-wide, Translocation Sequencing (HTGTS) to identify translocations.

"Current methods … are just providing a value of editing from the treatment and providing a value of editing from the control," Hendel said. This approach can generate false negatives.

The researchers said that CRISPector improves on earlier methods in four ways: increased accuracy of treatment versus control, thanks to the statistical modeling; the ability to identify alternative cut sites when an initial analysis is off target; detection of adverse structural variations and translocations in editing experiments from analyzing multiplex PCR and NGS data; and reporting of statistical confidence intervals for suspected off-target activities.

Most of the previous CRISPR quality-control methods are not suitable for multiplex PCR experiments, the authors explained in a Nature blog post that accompanied the paper. While CRISPector was designed for multiplex PCR environments, it also can support singleplex PCR.

Hendel said that the open-source CRISPector software is suitable for both academic labs conducting preclinical studies and for biotech companies looking to evaluate CRISPR processes. It is freely available on Github for academic use now, and he said that the company is looking at several options for commercializing the technology.

Coralville, Iowa-based IDT referred questions about commercialization of CRISPector to Hendel.

The paper's authors said that a better understanding of potential off-target activities in CRISPR gene editing would help geneticists design better editing strategies.

The authors said that researchers at the Hendel Lab at Bar-Ilan University and the Yakhini Research Group at Technion hope that CRISPector will help them explore new treatments for genetic disorders of the immune system as well novel immunotherapies for cancer.

They also said that future versions of the software will support the use of unique molecular identifiers as part of multiplex PCR experiments in an effort to minimize amplification-related biases.

"One of the challenging things in the CRISPR world is that it generates different types of adverse events, and it's difficult to pick all the different types of adverse events by applying one assay," Hendel said.

While his team is working with multiplex PCR, it also is looking to develop new assays and adapt existing assays to the CRISPector statistical model.

"There are a number of types of damage that the CRISPR is doing," he said. "To actually address each of them, it might be that you need a different type of experiment, an assay, to evaluate it," Hendel speculated.