Quantifying Editing Efficiency and Off-Target Events

Every editing experiment produces three things that must be reported:

1. Efficiency — the fraction of alleles carrying the intended edit.
2. Product spectrum — what else is present: indels, larger deletions, rearrangements.
3. Off-target activity — edits at unintended sites.

Most publications report the first approximately, the second rarely, and the third selectively. Each omission has a cost: efficiency without a spectrum hides mixed populations; a spectrum without off-target data cannot support a specificity claim.

For publication-grade work, amplicon NGS is the default. Sanger-plus-deconvolution is acceptable for screening clones before committing to sequencing, provided the final claim rests on sequencing.

An indel frequency of 70% does not mean 70% null alleles. The spectrum matters:

• • Frameshift versus in-frame indels. A 3-bp deletion in-frame may leave a partially functional protein — the most common reason a "knockout" is not a knockout.
• • Large deletions. Some edits produce kilobase-scale deletions that amplicon PCR may not amplify, hiding the true allele.
• • Allelic heterogeneity. In diploid cells, two different indels are common; each may have a different consequence.
• • Wild-type contamination. A mixed population reported as a clone is a frequent error in pool-based experiments.

Reporting a spectrum — the top alleles with frequencies — takes one supplementary figure and prevents most challenges to a knockout claim.

For base and prime editing, which do not create double-strand breaks, break-based methods are poorly suited; RNA-guided off-target activity is better assessed by targeted amplicon sequencing at predicted sites plus, where the claim is strong, whole-genome sequencing at adequate depth.

A common standard: predicted sites for the guide, ranked by similarity, with amplicon sequencing across the top 10–20 and a stated detection limit.

The fifth row is where claims usually weaken: "no off-target effects" needs a number, a site count, and a detection limit.

• • Reporting efficiency from a pool but claiming a clone. Pools and clones have different allele distributions.
• • Ignoring in-frame indels. Or confirming protein loss only by mRNA.
• • Testing off-targets at sites chosen after seeing the phenotype. Pre-specify the site list.
• • Using the wrong method for the editor. DSB-based off-target methods for base editors.
• • Assuming copy number. Tandem integration is common in knock-ins and invisible to junction PCR.

Not every project needs the full matrix. Three tiers cover most situations:

The middle tier is what most publications should meet, and it is achievable within a normal budget: amplicon sequencing is the expensive step, and it replaces several cheaper assays rather than adding to them.

The third tier is where the word "specificity" becomes defensible. Everything below it supports a statement about the intended locus, not about the genome.

EDITGENE reports per-clone amplicon sequencing, allele spectra, zygosity by digital PCR, off-target panels at predicted sites with stated detection limits, and copy-number confirmation — for knockout, knock-in, and base- or prime-edited lines.

Tell us the edit and the claim. Contact EDITGENE to request a project consultation or quote.

Efficiency and off-target rates vary by locus, editor, delivery, and cell type; the methods here define what can be measured, not expected values.

A specificity claim is bounded by the sites examined and the detection limit of the assay used, and should be reported with both.

• • Tsai SQ, et al. GUIDE-seq enables genome-wide profiling of off-target cleavage by CRISPR-Cas nucleases. Nature Biotechnology. 2015;33:187–197.
• • Tsai SQ, et al. CIRCLE-seq: a highly sensitive in vitro screen for genome-wide CRISPR-Cas9 nuclease off-targets. Nature Methods. 2017;14:607–614.
• • Brinkman EK, et al. Easy quantitative assessment of genome editing by sequence trace decomposition. Nucleic Acids Research. 2014;42:e168.
• • Vouillot L, et al. Comparison of T7E1 and surveyor mismatch cleavage assays. G3. 2015;5:407–415.