Base Editing vs Prime Editing
Base editing and prime editing are both next-generation CRISPR refinements invented by David Liu's lab at the Broad Institute that avoid double-strand DNA breaks. Both achieved landmark clinical milestones in 2025: Beam Therapeutics demonstrated the first clinical correction of a point mutation (BEAM-302 for AATD, March 2025), while Prime Medicine published the first-in-human prime editing results in the New England Journal of Medicine (December 2025), showing 69-83% gene correction in CGD patients. Meanwhile, prime editing efficiency has been revolutionized — PE7 (Nature 2024) achieves 21-fold improvements, and AI-designed protein binders (Cell 2025) push efficiency 18.8x beyond PEmax, closing the historical gap with base editing.
Last updated: March 29, 2026
Base Editing
Uses a deaminase enzyme fused to a Cas9 nickase to chemically convert one DNA base into another without cutting the double helix. Two main types: CBE (cytosine base editor, C→T) and ABE (adenine base editor, A→G).
Prime Editing
Uses a reverse transcriptase fused to a Cas9 nickase plus a prime editing guide RNA (pegRNA) that directly templates the desired edit. A true 'search-and-replace' tool that makes virtually any small precise edit without bystander mutations or RNA off-targets.
Key Specifications
| Feature | Base Editing | Prime Editing |
|---|---|---|
| Mechanism | Chemical base conversion via deaminase (nickase, no DSB) | Reverse transcriptase + Cas9 nickase + pegRNA template |
| Precision | Single-base, but with bystander editing in ~4-8 nt window | Exact — no bystander editing, no RNA off-targets |
| Edit types | 4 transitions only: C→T, G→A (CBE) and A→G, T→C (ABE) | All 12 transitions/transversions + insertions + deletions |
| Insertions/Deletions | No — cannot perform | Yes (up to ~50 bp insert, ~80 bp delete) |
| Editing efficiency | High for supported edits (30-80%+) | PE7: 21x over PE2; PE7+MLH1-SB: 18.8x over PEmax; 40-70%+ at optimized loci |
| Off-target risk | Low indels, but significant RNA off-targets (ABE) and bystander edits | Very low — no deaminase, no bystander, no RNA off-targets |
| Bystander editing | 30-60% at neighboring bases in editing window | — |
| Protein size | ~5.2 kb | ~6.3 kb (standard); PE6a/b are AAV-compatible |
| Clinical stage | Phase 1/2, BLA track (BEAM-101 targeted YE 2026) | Phase 1/2 (PM359 CGD — NEJM-published, 69-83% correction) |
| Disease coverage | ~30% of known disease-causing mutations | ~89% of known disease-causing mutations |
| Key companies | Beam Therapeutics, Verve/Eli Lilly | — |
| Invented | 2016 (David Liu, Broad Institute) | 2019 (David Liu, Broad Institute) |
| Upcoming INDs | — | PM577 (Wilson's, H1 2026), AATD (mid-2026) |
| Key partnerships | — | Bristol Myers Squibb (CAR-T), CF Foundation ($39M) |
Base Editing
Advantages
- First clinical genetic correction: Beam's BEAM-302 corrected AATD Z-mutation in patients (March 2025)
- BEAM-101 for SCD: 30 patients dosed, FDA RMAT designation, BLA submission targeted year-end 2026
- High efficiency for supported transitions: 30-80%+ for C→T and A→G edits in most cell types
- More clinical programs running: ~8-10 active trials (Beam, Verve/Lilly, academic)
- Major pharma validation: Pfizer $300M + Eli Lilly $200M deals with Beam (2025)
- Smaller protein (~5.2 kb) — simpler delivery logistics than prime editor
- Verve's VERVE-102 showed 53-69% LDL-C reduction from single IV dose for cardiovascular disease
Limitations
- Fundamentally limited: only 4 of 12 possible base conversions (C→T, G→A, A→G, T→C)
- Cannot make insertions or deletions — excludes ~70% of disease-causing mutations entirely
- Bystander editing: converts ALL target bases in the ~4-8 nt window, creating unwanted mutations at 30-60% rates
- Transcriptome-wide RNA off-target deamination: ABE8e causes tens of thousands of A-to-I RNA edits
- Fixed editing window position relative to PAM — less flexible targeting than prime editing
- Verve's VERVE-101 safety signals led to program deprioritization — in vivo base editing carries risks
Prime Editing
Advantages
- NEJM-published clinical success: PM359 achieved 69% and 83% gene correction in two CGD patients by Day 30 (December 2025)
- Most versatile: all 12 base changes + insertions (up to ~50 bp) + deletions (up to ~80 bp)
- Corrects ~89% of known disease-causing mutations vs. ~30% for base editing
- Zero bystander editing — edits ONLY the exact targeted position, unlike base editing's window problem
- No RNA off-targets — no deaminase means no transcriptome-wide off-target editing
- Efficiency revolution: PE7 achieves 21.2-fold improvement (Nature 2024); PE7 + AI-designed MLH1-SB reaches 18.8x over PEmax (Cell 2025)
- PE6 variants (a-d): 2-20x more efficient, some small enough for AAV delivery — solving the size problem
- Multiple IND filings planned: Wilson's Disease (H1 2026), AATD (mid-2026); PM359 data may support accelerated BLA
Limitations
- Fewer clinical programs than base editing as of 2026 — still earlier in the regulatory pipeline
- Larger protein complex (~6.3 kb standard, though PE6a/b are smaller) — delivery historically more complex
- pegRNA design requires more optimization than base editing guide RNAs (improving with AI tools like DeepPrime)
- Prime Medicine winding down CGD program despite efficacy — market too small for commercial viability
- PM359 required myeloablative busulfan conditioning — complex ex vivo process
- In vivo prime editing efficiency in humans not yet demonstrated (preclinical LNP delivery ~10-20% in mouse liver)
The Verdict
Both technologies achieved landmark clinical milestones in 2025, but prime editing has emerged as the more powerful and precise tool. Base editing remains efficient for C→T and A→G transitions, with more clinical programs and a BLA-track therapy (BEAM-101). But it can only address ~30% of disease mutations, suffers from bystander editing (30-60% at neighboring bases), and causes transcriptome-wide RNA off-targets. Prime editing is the precision universal tool: it corrects ~89% of disease mutations with zero bystander editing and zero RNA off-targets. The historical efficiency gap has been dramatically closed — PE7 achieves 21-fold improvements (Nature 2024), and AI-enhanced PE7 reaches 18.8x over PEmax (Cell 2025). Prime Medicine's NEJM-published PM359 results (69-83% gene correction in CGD patients) prove that prime editing delivers extraordinary clinical efficacy. For the narrow set of C→T/A→G transitions in favorable editing windows, base editing remains a strong choice. For everything else — and for applications requiring precision without bystander damage — prime editing is the superior technology.
Sources & References
- [1]PM359 NEJM publication — first-in-human prime editing
- [2]Prime Medicine PM359 clinical data (69-83% correction)
- [3]PE7 — 21-fold efficiency improvement (Nature 2024)
- [4]AI-designed MLH1-SB + PE7: 18.8x over PEmax (Cell 2025)
- [5]PE6 evolved prime editors — smaller and more efficient (Broad Institute)
- [6]Beam BEAM-302 first clinical genetic correction
- [7]Beam Pfizer $300M + Lilly $200M deals
- [8]Prime editing emerging trends review (2025)
- [9]Prime Medicine FY2025 results & pipeline
- [10]Evolution of prime editing review (Advanced Science 2026)