A machine learning model just made prime editing dramatically easier to design

Key Findings

🔬 Six Genes Were Quietly Blocking Gene Editing the Whole Time

• A genome-wide CRISPR screen across 19,114 genes identified six proteins that act as negative regulators of nonviral gene delivery — meaning knocking them out makes editing work better, not worse.
• Depleting the top two hits, GJB2 and BET1L, improved base editing efficiency up to 6-fold across multiple cell types and gene targets. In a patient-derived model of retinal disease, lipid nanoparticle editing improved by over 3.5-fold — enough to restore ion channel function in a subset of cells.

⚠️ Off-Target Detection Got a Serious Upgrade

• UNCOVERseq, a new in-cell off-target nomination workflow, hit 97.6% analytical sensitivity and 78% precision when benchmarked against published methods — outperforming the field on both metrics.
• Tested across 192 guide RNAs in blood stem cells, it showed that off-target sites identified using double-strand break tools retain strong rank-order agreement with base editors, which only nick one strand. That means one nomination workflow can inform risk assessment across multiple editing platforms.

🧲 Magnetic Nanoparticles Outperformed Standard CRISPR Delivery by Over 5-Fold

• Swapping conventional lipid-based transfection for magnetic nanoparticle delivery (magnetofection) pushed average editing efficiency from roughly 8–12% up to 42–45% across both Cas9 and Cas12a systems — a 3- to 5-fold jump depending on the nuclease.
• The boost held across multiple cell lines and genomic targets, and also improved prime editing efficiency from 3.8% to nearly 14% on average. The effect appears to be about delivery, not the editor itself.

💉 A Bespoke Base-Editing Drug Was Made for One Baby — Here's What That Means

• In 2025, an infant with a severe urea cycle disorder received a lipid nanoparticle-delivered base-editing therapy designed specifically for their individual mutation — reportedly the first recipient of a customized in vivo gene-editing medicine.
• This review examines the clinical rationale, delivery mechanics, and ethical terrain of so-called N-of-1 genetic medicine, including questions of consent for non-autonomous newborns, cost, and whether a single case can anchor a scalable platform.

🦠 A Tumor Vaccine That Works in Cancers Immune Checkpoint Drugs Can't Touch

• Researchers used CRISPR-Cas9 to knock out TGF-β1 in tumor cells, then combined the modified cells with an immune-activating compound to create an off-the-shelf whole-tumor vaccine called Tβ1KO/PolyIC-Vac.
• In mouse models of mismatch repair-proficient tumors — the majority of solid cancers that don't respond to checkpoint inhibitors — the vaccine suppressed tumor growth, cut lung metastases, and synergized with anti-PD-1 therapy to raise complete response rates.

🧬 A New Cas9 Structural Feature Explains Why Some Guide Designs Work Better Than Others

• Cryo-EM and molecular dynamics revealed a previously unknown structural motif in Cas9 — the guide repeat clasp — that checks the shape of the guide RNA and coordinates with internal checkpoints before allowing DNA cutting to proceed.
• Comparing the natural two-part guide RNA to the widely used single-piece fusion version showed measurable differences in editing efficiency and off-target activity across 255 tested genomic sites, with no simple sequence rule predicting which format would win.