Gene editing has come a long way since the discovery of early programmable nucleases like zinc finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs), and more recently, CRISPR-Cas9. The versatility and ease of customization of CRISPR technology quickly gave rise to a diverse ecosystem of novel gene-editing modalities. These advancements include Cas nuclease variants such as Cas12a, as well as a growing suite of engineered nucleases with distinct and tunable properties, including altered protospacer adjacent motif (PAM) requirements, enhanced efficacy, and modified cleavage behavior.
Base and prime editors have further expanded these capabilities by enabling targeted single nucleotide substitutions, insertions, and deletions with greater flexibility and accuracy. As a result, the CRISPR toolbox has grown at remarkable speed, supporting therapeutic strategies across oncology, metabolic disease, and a wide range of rare genetic disorders.
Despite the rising potential of gene editing tools, their underlying mechanism of action remains central to both their therapeutic promise and risk profile. Traditional Cas nucleases function by introducing double-strand DNA breaks (DSBs) at defined genomic loci, relying on the cell’s endogenous DNA repair pathways to disrupt or modify specific genes. This strategy remains the most widely used form of gene editing and offers the potential for permanent genomic change, raising the prospect of durable and, in some cases, curative, single-treatment therapies.
However, unintended insertions or deletions (indels) at both the on-target and off-target sites in the genome caused by promiscuous nuclease activity are an important concern given their potential to cause permanent genomic damage. Newer modalities such as base editing and prime editing were developed to enable more precise genetic modifications without creating DSBs, expanding both the technical possibilities and the potential safety profile of genome engineering. Nevertheless, these supposedly DSB-free approaches have also been shown to induce unintended mutations and genotoxicity, specific to their mechanisms of gene editing.
The effects of CRISPR-Cas off-target activity
CRISPR-Cas systems are highly efficient and versatile, but unintended editing at sites beyond the intended target remains a significant safety concern. Off-target activity (OTA) arises when the CRISPR-Cas complex binds and cleaves DNA at loci other than the on-target site. The primary determinant of OTA is DNA sequence, where the number and position of mismatches and bulges at the off-target site can alter binding by the guide RNA (gRNA). OTA is also influenced by a range of other factors, including modifications to the nuclease or gRNA that alter specificity, genomic features such as chromatin accessibility and epigenetic state, and experimental variables including cell type, editing dosage, and delivery method.
This risk of OTA is further compounded by the wide range of possible editing outcomes and their consequences. Indels introduced at off-target sites are not always consistent across cell and tissue types, and can result in frameshift, nonsense, or missense mutations that disrupt gene regulation and expression. In severe cases, these alterations could drive oncogenic transformation if tumor suppressor genes or proto-oncogenes are affected.
Of even greater concern are large-scale structural variants, including translocations, deletions, duplications, inversions, chromosomal loss, and unintended integration of homology-directed repair donors. These can arise at both on- and off-target sites and represent a form of genotoxicity that bulk sequencing approaches may fail to detect entirely. Reflecting these risks, the FDA’s April 2026 guidance recommends next generation sequencing-based chromosomal integrity analysis for genome editors known to create DSBs, with specific assessment of translocation events between on- and off-target edit sites.
Most importantly, the FDA’s draft guidance now provides specific recommendations for NGS-based off-target analysis, requiring sponsors to use sequencing depth sufficient to detect low-frequency off-target events, employ strategies to minimize PCR amplification bias, and use human cells representative of the intended therapeutic target for both ex vivo and in vivo products. Market expectations are also shifting toward earlier, more systematic assessment of both on- and off-target activity throughout development to ensure that editing events are characterized earlier, mitigating the risk of genomic instability and malignant transformation in gene-edited cells.
Where off-target screening is falling behind
Despite the rapid advance of gene editing modalities, many teams still lack a reliable, sensitive, cell-based method to measure on- and off-target effects early enough to influence program decisions. As a result, gene-editing programs often rely on legacy methods that infer nuclease activity indirectly. For example, by sequencing DNA tags integrated at break sites, mapping the accumulation of DNA damage-response proteins, or profiling cleavage events in isolated genomic DNA exposed to nucleases in vitro , rather than directly quantifying edited alleles in their native cellular context. Because these approaches measure proxies of editing activity rather than the breaks themselves, they offer limited insight into the mechanisms driving off-target effects or how editing behavior might be modified to reduce them.
This is now more relevant than ever because the complexity of gene editing modalities and their mechanism of action has grown. Earlier editing platforms, such as ZFNs and TALENs, rely on comparatively large, protein- and amino acid-based recognition domains, imposing target binding constraints that have more predictable specificity profiles. CRISPR-Cas systems, however, introduce RNA-guided targeting, which allows for far greater flexibility in targeting, but also introduces a distinct mode of OTA. Cas nucleases tolerate more mismatches between the gRNA and target DNA than earlier editing systems, enabling cleavage at partially complementary genomic sites that can be difficult to predict.
Beyond Cas nucleases, the growing adoption of base editors and prime editors introduces additional off-target mechanisms, including off-target DNA nicking and bystander DNA deamination, that operate outside the DSB framework entirely, and for which most legacy detection methods were not designed. The FDA has explicitly recognized this challenge, noting in its April 2026 guidance that assays designed to detect DSBs may not be suitable for detecting the single-strand nicks created by base editors, and that sponsors should consider their editor’s mechanism of action when selecting an off-target analysis approach.
In vitro biochemical approaches offer sensitivity and an unbiased assessment of potential break sites, but they identify these sites in the absence of cellular context. Chromatin architecture, epigenetic state, and cell-type-specific repair pathway activity all influence where and how often editing occurs — factors that are not replicated in a cell-free system.
Legacy cell-based methods partially address this by operating in live cells but depend on the efficient integration of an exogenous tag molecule that is cytotoxic to many primary cell types, restricting their applicability precisely where therapeutic relevance is highest. Additionally, the complex, PCR-based workflows used by most legacy methods introduce amplification bias, distorting the true frequency of break events. This further complicates interpretation and can make it difficult to detect rare off-target against the backdrop of PCR-induced noise. Collectively, these limitations leave significant blind spots in off-target profiles and make legacy approaches poorly suited for iterative OTA assessment across the full range of modern editing modalities.
Why speed matters, beyond efficiency
Slow, indirect, and disconnected data are poorly suited to the iterative, high-confidence decision making that modern gene editing programs require. Inadequate characterization of off-target editing creates significant regulatory and financial risk and may lead to late-stage clinical holds, program termination, and loss of investor confidence.
In addition to the technical limitations of individual methods, legacy workflows also share operational weaknesses characterized by long turnaround times measured in weeks or months, which can delay decisions around guide selection or dose optimization, and complex on- and off-target analysis workflows, resulting in ambiguous data and reduced reproducibility. The speed at which complete off-target data can be generated has a direct bearing on patient outcomes and should be treated as a critical variable and considered as early as possible in gene editing workflows. Key program decisions, including final gRNA sequence and when to progress into preclinical or clinical work, depend on timely and complete off-target insights.
In the early research and development stages, companies often test hundreds or thousands of gRNAs, primarily focusing on editing efficacy, while safety assessment is evaluated more extensively only in later stages as candidate programs are narrowed. This approach risks advancing candidates with promising early efficacy but incompletely characterized safety profiles into later development stages, where unanticipated off-target liabilities are harder and more costly to resolve. Furthermore, many candidates that are discarded early on the basis of efficacy alone may have excellent specificity profiles that would make them particularly well suited for use in a patient.
Identifying safety issues late during development comes with risk and the potential for substantial redesigning of editing components, and can even lead to program termination, all of which extend timelines to reach patients. Delayed off-target data also misses the opportunity for data-driven cycles of learning during development, a central principle used to advance CRISPR-based gene-editing modalities. Meaningful comparison of editor variants or guide designs becomes difficult when off-target analysis cannot keep pace with candidate screening and ultimately slows therapeutic development.
Comprehensive and rapid genome-wide characterization of on- and off-target DNA breaks induced by gene editing would therefore help de-risk programs during discovery and preclinical phases of development, before assets progress toward clinical submission.
What the future of off-target screening looks like
Addressing the above challenges will require solutions that move beyond indirect measurements and cell-free assays toward rapid, accurate, genome-wide detection of editing outcomes within relevant cellular contexts. As gene editing programs evolve, the demands on the quality and quantity of off-target data will continue to increase, raising the bar for what screening approaches must deliver to support safe and effective therapies.
A key requirement is a genome-wide, unbiased approach that does not rely on predefined predictions of potential off-target sites. For novel editing modalities whose activity is still being characterized, an approach that makes no prior assumption about where editing has occurred is the only one that can provide comprehensive insights. That is increasingly what regulators recommend, and what development teams need to build confidence in their candidates. The FDA’s guidance now requires that off-target editing and chromosomal translocation assessments be completed prior to Investigational New Drug (IND) submission, with early engagement through pre-IND meetings encouraged, effectively establishing early, comprehensive characterization as a regulatory expectation.
Equally important is the ability to measure editing events directly within intact, biologically relevant cells, rather than inferring them from proxy signals or cell-free cleavage patterns. Cell-based, in situ approaches provide a more faithful representation of how editing occurs in therapeutic contexts and generate data that is both more accurate and more actionable.
To ensure quantitative reliability, these methods must also avoid amplification-induced distortion. PCR-based workflows introduce bias that can skew the relative frequency of editing events, increase background noise, and reduce the precision needed for confident decision-making, particularly when detecting rare off-target events.
Finally, integrating on- and off-target characterization into a single workflow enables a more complete view of editing behavior. When both are assessed from the same cells and sequencing run, variability is reduced and data interpretation becomes more coherent. If delivered in-house with turnaround times measured in days rather than weeks or months, such workflows also allow teams to retain control over data and intellectual property while generating insights early enough to meaningfully influence program direction.
Looking ahead
Legacy off-target screening approaches are failing to keep up with expanding gene editing modalities, therapeutic pipelines, and regulatory expectations. As gene editing moves across an increasingly broad range of disease areas, the ability to comprehensively and accurately measure off-target effects has become critical to both development and regulatory decision-making. Modern approaches encompassing genome-wide, cell-based, PCR-free characterization, and that are fast enough to return results in days, directly address these challenges (Table 1).
Table 1: Comparative overview of methods for off-target activity characterization
Feature | Cellular methods (e.g., GUIDE-seq) | Cell-free, in vitro biochemical methods (e.g., SITE-seq) | Genome-wide, in situ methods (e.g., INDUCE-seq) |
Cell-based | Yes | No | Yes |
PCR-free | No | Varies | Yes |
Endogenous break detection | Limited | No | Yes |
Compatible with primary cells | Limited | N/A | Yes |
Genome-wide, unbiased | Partial | In vitro only | Yes |
In-house, standardized workflow | Complex | Complex | Yes |
Results turnaround | Weeks to months | Weeks to months | Days |
Integrated bioinformatics | No | No | Yes |
Identifying off-target liabilities early, before candidates progress into costly preclinical and clinical stages, means fewer late-stage failures and less time spent resolving problems that earlier data would have prevented. Guide designs can be iterated quickly, the strongest candidates selected with confidence, and regulatory packages built on robust, empirical evidence from the outset.
Programs that are better characterized earlier move through development with less friction, fewer unexpected findings, fewer demands for additional studies, and greater confidence from regulators and investors alike. This in turn allows gene-edited therapies to reach patients safely, without unnecessary delays.
