Gene Editing Weekly — August 22–28, 2026: PBGENE-DMD enters the clinic as prime assembly stretches the payload envelope

The coverage window is August 22–28, 2026, Eastern Time. Four developments cleared the bar for this issue: Precision BioSciences started dosing PBGENE-DMD in Duchenne muscular dystrophy; Nature Biotechnology published a prime-editing method for large DNA replacement; a bioRxiv preprint benchmarked machine-learning classifiers for CRISPR-Cas9 off-target sites; and the American Society of Gene & Cell Therapy (ASGCT) asked HHS's Office of Inspector General (OIG) for clearer rules on participant support in cell and gene therapy trials.

Signal	What changed this week	Evidence boundary
Clinical	PBGENE-DMD reached first-patient dosing in the Phase 1/2 FUNCTION-DMD trial on August 24. 1	Trial start; efficacy and safety data are still pending.
Platform research	Prime assembly reported up to 57.8% replacement efficiency with a 2.9-kb donor in HEK293T cells. 2	Peer-reviewed paper with preclinical cell and mouse evidence.
Computational safety	A preprint compared five classifiers on a GUIDE-seq off-target dataset with 54 validated cleavage sites among 95,829 candidates. 3	Preprint; sequence-based benchmark rather than clinical validation.
Policy	ASGCT submitted comments to HHS OIG on the Anti-Kickback Statute and Beneficiary Inducements Civil Monetary Penalty framework. 4	Advocacy submission; OIG has yet to state an outcome here.

Clinical development

PBGENE-DMD moves from pipeline story to patient dosing

Precision BioSciences announced on August 24 that the first patient received PBGENE-DMD at Arkansas Children's Hospital in the Phase 1/2 FUNCTION-DMD trial. The filing makes this a new clinical milestone after earlier pipeline coverage: the program has moved from development planning into human dosing. 1

FUNCTION-DMD is enrolling ambulatory children ages 2–7 with DMD mutations between exons 45 and 55. Precision BioSciences estimates that this genotype band covers up to 60% of people with Duchenne muscular dystrophy. PBGENE-DMD is designed to edit the patient's own dystrophin gene in vivo, with the aim of producing a near-full-length functional dystrophin protein from the endogenous locus. 5

The development signal has two separate parts. The first is operational: a specialized pediatric site has dosed the first participant. The second is biological: the program is pursuing permanent editing of the native gene rather than delivery of a shortened replacement gene. The first part is established now; the second remains a therapeutic hypothesis until the trial produces biomarker, functional, and safety data. Precision BioSciences expects initial safety data by the end of 2026. 5

For researchers, the next readout to watch is the relationship among delivery, editing at the intended DMD region, immune findings, and dystrophin production. For investors, first-patient dosing establishes development cadence and cohort design; it supplies no estimate of efficacy or probability of approval.

Peer-reviewed platform research

Prime assembly targets kilobase-scale replacement without the paper's DSB-centered route

A team led by Hojun Jung and Sangsu Bae at Seoul National University published "Precise genomic integration of large DNA fragments by donor-directed annealing using prime editing" online in Nature Biotechnology on August 28. The paper introduces prime assembly (PA), which uses prime editors to create complementary 3-prime DNA flaps on the genomic target and a donor molecule. The flaps anneal, allowing the donor sequence to be copied into the target region. 2

The authors describe three architectures: single-flap PA, single-flap PA with an additional nick, and dual-flap PA. The system accepted plasmid and linear double-stranded DNA donors from 1.0 to 6.5 kb. In HEK293T cells, dual-flap PA reached 57.8% replacement efficiency with a 2.9-kb donor, while integrated PA fragments showed reported accuracy above 90%. 2

Figure 1 from the prime-assembly paper, showing three architectures and initial efficiency experiments — *Nature Biotechnology Figure 1 compares single-flap PA, single-flap PA with a nick, and dual-flap PA, then shows how flap spacing affected the initial HEK293T experiments. 2*

The paper also tested PA in primary human T cells, where site-specific chimeric antigen receptor integration reached 28.1%. A hydrodynamic-injection experiment in mice measured an average integration efficiency of 4.3% in GFP-positive hepatocytes. Those experiments broaden the platform's scope across cell types and an in vivo model, while keeping the evidence preclinical. 2

The practical question is how these percentages behave in the settings that matter for a therapeutic program. The paper reports cell-line, primary-T-cell, and mouse experiments; it does not establish a clinical delivery system, disease correction, durability, or a human safety profile. The next useful data would include reproducibility across disease-relevant loci, payload sequence classes, primary patient cells, and genome-wide assessment of unintended integration and rearrangement.

Preclinical and computational evidence

The off-target benchmark puts precision-recall ahead of an impressive ROC number

Deeksha Sarvi and Juhitha Alasyam posted a bioRxiv preprint on August 23 titled "Classifying CRISPR-Cas9 Off-Target Cleavage Sites from GUIDE-seq Data: A Class-Imbalanced Machine Learning Benchmark." The authors tested five classifiers on a published GUIDE-seq dataset containing 95,829 candidate off-target sites for five sgRNAs. Only 54 sites, or 0.06%, had experimental validation as true cleavage sites. The record is a preprint and has not been certified by peer review. 3 6

Gradient boosting on combined mismatch and sequence features produced the strongest reported held-out result: ROC-AUC 0.997, PR-AUC 0.355, and best F1 0.50. The precision-recall area under the curve (PR-AUC) matters more here because the positive class is extremely rare. A model can achieve a high receiver-operating-characteristic area under the curve (ROC-AUC) while still producing too many false positives for a rare-event screening workflow. 3

The benchmark also found that validated cleavage sites carried fewer mismatches across the 23-nucleotide target than non-cleaved candidates: a mean of 3.6 versus 5.9 mismatches. The PAM-proximal seed region and the PAM itself showed lower mismatch rates among validated sites. These observations fit established Cas9 target-recognition patterns, while the small number of positive sites limits how far the result can be generalized. 3

For a translational safety workflow, the useful takeaway is a ranking tool for follow-up experiments. The preprint supplies a benchmark on one dataset and a held-out split; it does not replace GUIDE-seq, targeted deep sequencing, orthogonal assays, or product-specific risk assessment. Independent datasets, diverse nuclease and guide contexts, and experimentally confirmed positives remain the next tests.

Policy and governance

ASGCT asks OIG to clarify support for trial participants

On August 24, ASGCT submitted comments to HHS OIG's request for information on the federal Anti-Kickback Statute and the Beneficiary Inducements Civil Monetary Penalty provisions as they apply to clinical-trial participant remuneration. ASGCT asks OIG to consider a safe harbor or corresponding exception for support reasonably necessary to take part in an FDA-regulated clinical trial. 4

The requested support covers travel, lodging, meals, parking, local transportation, medically necessary transportation, caregiver time, childcare or dependent care, and some protocol-related or temporary supportive services. ASGCT's argument rests on the operating model of cell and gene therapy trials: specialized centers can leave patients geographically dispersed, administration can require intensive short-term monitoring, and long-term follow-up can continue for years. 4

ASGCT also asks OIG to distinguish support that removes a participation barrier from payment that induces the use of unrelated federally reimbursable services. The society proposes predefined sponsor policies, documentation, consistent administration, and continued Institutional Review Board review of informed consent and undue influence. The submission favors support based on participation burden rather than income tests, sponsor type, or a conventional Phase 1/2/3 label. 4

The current status is a policy submission. OIG's response, the form of any guidance, and whether any safe harbor or exception is proposed remain the watchpoints. Trial sponsors and sites should treat the submission as a signal about operational friction and representativeness, rather than as a change in permissible payment rules.

Regulatory watch

The FDA's Cellular & Gene Therapy Guidances page lists two entries under 8/2026: a frequently asked questions document on developing potential cellular and gene therapy products, and a draft guidance on potency assessment of active immunotherapy products. The page gives month-level dates for those entries, so the current-week review cannot assign either entry specifically to August 22–28. 7

The current-week product-specific regulatory ledger therefore remains empty in this review: no new FDA approval, complete response letter, advisory-committee action, IND clearance, or BLA/NDA acceptance was verified for a gene-editing product within the window. The month-only FDA dates are a reporting limit, rather than evidence of regulatory inactivity. The two August guidance entries remain useful watch items for sponsors tracking CBER expectations.

Issue summary and next watchpoints

This week's four items sit at different points on the path from platform promise to clinical use. PBGENE-DMD supplies the human dosing milestone. Prime assembly expands the preclinical design space for large replacement payloads. The off-target benchmark tests how a safety workflow might prioritize rare cleavage events. ASGCT's submission addresses whether patients can realistically enter and remain in the trials that generate the evidence. 1 2 3 4

The next research and investment work is concrete:

Track PBGENE-DMD's first safety and biomarker data, expected by the end of 2026, and keep the first-patient event separate from efficacy evidence. 5
Test whether prime assembly's reported efficiencies and accuracy persist in disease-relevant primary cells, larger or more difficult payloads, and broader off-target integration surveys. 2
Treat the off-target classifier as a prioritization hypothesis until independent datasets and orthogonal experiments reproduce its precision-recall behavior. 3
Watch for OIG's response to participant-support comments and any FDA clarification that resolves the August guidance dates. 4 7