What Must Be Measured, and What Can Be Shared: NGS, prior knowledge, and genome-editing safety in three 2026 FDA publications

In 2026, FDA issued three documents that together answer two practical questions for genome-editing programs: what genomic safety evidence should be generated before an IND, and what evidence can be reused across products. This summary is written to help ourselves, and hopefully the community, to digest them a bit easier and faster. The three documents are:

• April 2026 draft: Safety Assessment of Genome Editing in Human Gene Therapy Products Using Next-Generation Sequencing (OTP/CBER; Docket #: FDA-2026-D-1255).
• June 2026 draft: Leveraging Prior Knowledge in the Development of Human Gene Therapy Products Incorporating Genome Editing (OTP/CBER; Docket #: FDA-2026-D-1257).
• August 2026 final: Frequently Asked Questions — Developing Potential Cellular and Gene Therapy Products (OTP/CBER; Docket #: FDA-2024-D-4311).

All three are recommendations, not binding requirements. The April and June documents remain drafts, not for implementation. The August FAQ is final. Read together, they describe a sequence: measure first, document the method well enough that parts of it can be shared, and complete the product-specific work before the original IND.

They do not lower the evidentiary bar. They specify it.

1. What the April draft asks sponsors to measure

The April draft is the measurement document. It applies to DNA editors, epigenetic editors, and RNA-cleaving editors, used ex vivo or in vivo. It sits on top of the January 2024 genome-editing guidance. The safety questions it pointed out are off-target editing and loss of genome integrity.

The operational recommendations are specific:

• Nominate off-target sites and confirm them. Nomination should use more than one approach (cell-based, biochemical, in silico, and/or NGS), matched to the editor’s mechanism of action. Confirmation is targeted testing at nominated sites. A subset of nominated sites may be confirmed if the selection criteria are justified; stringent filtering that simply drops sites is discouraged.
• Profile the on-target site for intended and unintended edits. Report the on-target edit rate. Samples used for on- and off-target analysis should be edited so that on-target rates are comparable to the proposed drug product.
• Match read length to the biology. Short-read methods may be adequate for changes of about 50 bp or less. Long-read methods are recommended when the modality creates double-strand breaks or can produce large insertions or deletions.
• Sequence deep enough to detect events that occur at lower frequency than the on-target edit. Support that sensitivity with in-house engineering-run data and/or peer-reviewed literature. Minimize PCR and primer bias where applicable.
• For modalities that create double-strand breaks or medium-to-large indels, perform a sensitive, quantitative NGS assessment of chromosomal integrity, including translocation rates involving on-target sites. If confirmed off-target sites exist, also assess translocations between on-target and off-target sites.
• Account for human genetic variation in in silico nomination (database choice, population stratification, common and rare variants, including variants that create a PAM (protospacer adjacent motif) or reduce mismatches).
• Complete the recommended off-target and chromosomal-translocation studies before the original IND. INTERACT packages can be brief (editor, mechanism, planned methods). Pre-IND packages should include sequencing strategy, data-quality and depth criteria, samples, and completed study reports if available.

The draft also specifies what a usable NGS report contains: cell type and input amount, library and primer details, bioinformatics tools and command-line records, reference sequences, acceptance criteria for quality, depth, and alignment, and tabulated results. Nominated and confirmed off-target sites should be annotated (coordinates, mismatches/bulges, PAM, genic context, and a risk discussion that may use prior knowledge).

None of this is a single assay. Nomination, confirmation, on-target characterization, and chromosomal integrity are different questions. They require different methods.

2. What prior knowledge can, and cannot, replace

The June draft is the reuse document. Leveraging means using public knowledge or platform knowledge to fully or partially avoid generating new data. Public knowledge is generally accepted scientific knowledge. Platform knowledge is what a sponsor has learned from similar products and processes. A “designated platform technology” designation is not required.

On NGS specifically, FDA draws a useful line: methods can be shared more easily than product-specific results.

What can often be shared

• Off-target nomination and confirmation strategies across products that use editors with the same mechanism of action.
• A genomic-integrity method used for one product, applied to a similar assessment of another.
• Assay parameters (input amount, editor concentrations, culture methods, controls) when intended on-target edit rates are similar.
• An NGS strategy, including confirmatory testing at a defined depth and quality, and the same class of method for nomination, on-target assessment, or chromosomal integrity.
• Bioinformatics pipelines used on the same assay type. Target sequence and mismatch parameters still have to be reset. A GUIDE-seq pipeline is the example in the draft: the pipeline can be reused; the input target sequence cannot.
• Off-target data, and in some cases genomic-integrity data, only when the gRNA, editor, and cell type are identical to another program. Extra chromosomal-integrity work may still be needed if patient cells have impaired DNA repair.

What generally cannot be shared

• Off-target assessment data across products that use different gRNAs or sequence-recognition components.
• On-target editing-outcome data across different genomic loci, even with the same editing mechanism.
• Genomic-integrity results when products edit different genomic locations.

That line is the point of the 2026 package. Method reuse is allowed where the science is actually the same. It is not a shortcut around a different guide, a different locus, or an unmeasured structural outcome.

3. What the CGT FAQ adds

The August FAQ is not an NGS methods document. It is a development-process document. Several answers still matter for how the April draft is used.

• Characterization is not release testing. Characterization describes the product. Release testing decides whether a lot is acceptable. Characterization assays need not be qualified, have acceptance criteria, or appear on the certificate of analysis. They should still be recorded and submitted in Module 3 at the appropriate phase. FDA recommends characterizing both drug substance and drug product.
• Analytical methods for first-in-human studies should be shown to be fit for purpose and phase-appropriate, with an evaluation of accuracy, precision, sensitivity, and specificity. Safety-related assays should perform adequately under actual conditions of use even at the initial IND.
• INTERACT is early, once a sponsor has identified a product and a development path. Pre-IND is later and more complete. The April NGS draft maps onto that split: strategy at INTERACT, methods plus data at pre-IND.
• Biodistribution for gene therapy products should use a quantitative, sensitive method (qPCR or ddPCR for vectors; LC-MS/MS for LNP products). Positive tissues should be followed with transgene RNA and/or protein. For in vivo editors, those tissues are also where additional on- and off-target work may be needed, as the April draft already notes.
• Genome-editing and integrating products remain in the long-term follow-up category. FDA continues to advise observation for delayed adverse events for as long as 15 years, because of cancer risk and other off-target effects.

The practical implication is narrow. Genomic safety NGS is characterization of the product’s editing profile, not a lot-release test. “Phase-appropriate” does not mean insensitive. The method still has to detect the events the April draft says matter, at the frequencies that matter.

4. How the three documents fit

Read as one package:

• The NGS draft tells sponsors what the genomic safety package should contain, in what samples, at what depth, and in what report format, before the original IND.
• The prior-knowledge draft tells sponsors which pieces of that package are platform (methods, pipelines, parameters) and which are product-specific (this gRNA, this locus, this cell type).
• The FAQ tells sponsors when to talk to FDA, what “fit for purpose” means, and how characterization differs from release.

The order is not optional. You cannot leverage a method you have not defined, and you cannot treat another product’s off-target list as yours because the editor class is the same. Measure first. Then argue what can be shared.

5. Independent, method-matched NGS

An independent laboratory sits in the middle of this, not as a substitute for the sponsor’s argument, but as a place where the methods can be run, documented, and where the June draft allows it, reused. The April draft’s questions map onto distinct NGS designs:

• Genome-wide off-target nomination in cells, for nuclease modalities: cell-based methods such as GUIDE-seq.
• Confirmation at nominated sites, and on-target small-variant quantification: targeted amplicon sequencing, with unique molecular identifiers (UMIs) to reduce PCR bias and support low-frequency detection.
• On-target large insertions, deletions, or transgene structure: long-read sequencing.
• Chromosomal translocations: targeted or primer-based methods, or other genome-wide NGS, with UMIs where quantification matters.
• Unintended integration and transgene integrity, when the delivery system or editing strategy makes those events relevant: genome-wide integration-site methods and long-read integrity analysis.

That is the design logic behind GeneGoCell’s Genome Editing Verification (GEV℠) platform: G-GUIDE℠ for genome-wide off-target identification; G-Amp℠ for UMI-based on- and off-target confirmation; G-Trans℠ for translocation analysis; G-Int℠ for integration site, efficiency, and transgene integrity. The point is not that one platform “satisfies” a draft guidance. The point is that the questions in the April draft are method-specific, and the June draft only lets the methods be shared when they are the same methods, documented the same way.

Two reporting habits in the April draft are easy to underweight and expensive to retrofit: acceptance criteria for quality, depth, and alignment, and a clear account of what the assay can and cannot detect. Those are also what make a result usable later, by the same sponsor, on a related product.

Closing

These three documents published by FDA in 2026 did not ask sponsors to generate less evidence for genome-editing products. It asked them to generate the right evidence, in the right samples, at a depth that can see rare events, with a report that another reviewer or another program can actually use. What must be measured is still product-specific: this editor, this guide, this locus, this cell type. What can be shared is the method, the pipeline, and, narrowly, data from an identical edit. Independent verification does not change that split. It makes the split auditable.

References

1. U.S. Food and Drug Administration. Safety Assessment of Genome Editing in Human Gene Therapy Products Using Next-Generation Sequencing. Draft guidance for industry. April 2026. Docket #: FDA-2026-D-1255. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/safety-assessment-genome-editing-human-gene-therapy-products-using-next-generation-sequencing

2. U.S. Food and Drug Administration. Leveraging Prior Knowledge in the Development of Human Gene Therapy Products Incorporating Genome Editing. Draft guidance for industry. June 2026. Docket #; FDA-2026-D-1257. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/leveraging-prior-knowledge-development-human-gene-therapy-products-incorporating-genome-editing

3. U.S. Food and Drug Administration. Frequently Asked Questions — Developing Potential Cellular and Gene Therapy Products. Final guidance for industry. August 2026. Docket FDA-2024-D-4311. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/frequently-asked-questions-developing-potential-cellular-and-gene-therapy-products

Note. The April and June documents are drafts and, until finalized, represent FDA’s proposed thinking only. This Insight is a scientific reading of the three documents. It is not regulatory advice and does not claim that any assay package satisfies FDA expectations for a specific product.