Guide-seq in CRISPR: What It Measures That Prediction Misses
GUIDE-seq as an experimental CRISPR off-target detection method is a cell-based assay that catalogs Cas-induced double-strand breaks by capturing an end-protected double-stranded oligodeoxynucleotide tag and sequencing the adjacent DNA. It is used to map breaks that happened in those cells, including sites a computational predictor did not nominate. It is not a second definition of off-target effects, and it is not a software score.
What GUIDE-seq Measures
Tsai et al. (2015) named the method genome-wide, unbiased identification of DSBs enabled by sequencing. The measurement is a catalog of breaks that accepted the tag — on-target and off-target — without starting from a predicted list. They applied it to 13 RNA-guided nucleases in two human cell lines. An open copy of that paper is on PMC4320685.
What counts as an off-target effect is already defined on the CRISPR off-target effects page. This page does not rewrite that definition. It is the method job that page points to in one paragraph.
How a Tagged Oligo Marks Breaks the Predictor Never Listed
The Tsai abstract states the principle: GUIDE-seq relies on capture of double-stranded oligodeoxynucleotides into DSBs. The 2021 Nature Protocols GUIDE-seq paper restates the same move with the later working details: an end-protected dsODN tag is integrated at nuclease-induced breaks; tag-containing genomic DNA is then amplified and sequenced. That protocol, including cell culture, is timed at nine days; once tagged genomic DNA is in hand, library preparation, sequencing, and analysis are timed at three days. Those clocks are the authors' workflow, not a promise that every lab will finish on the same calendar.
Because the assay starts from a tag in a break, it can recover a site no mismatch model had listed. Because it needs the tag delivered into those cells, it is not a genome-wide certificate for every cell type. This section is a mechanism sketch, not a protocol reprint. Cell types that do not tolerate the oligo, and biochemical alternatives, belong to that protocol literature.
Prediction Versus GUIDE-seq
Prediction and GUIDE-seq are two jobs. Targeted sequencing of nominated sites is a third. Tsai compared GUIDE-seq with the computational methods then available and with ChIP-seq; the majority of identified sites were not detected by either. The 2021 protocol also notes that complementary biochemical methods can still recover sites a cell-based tag assay misses. Do not crown any column complete.
| Computational prediction | GUIDE-seq | Targeted sequencing of nominated sites | |
|---|---|---|---|
| Starting point | Spacer, nuclease, and a chosen genome under a mismatch-and-PAM model | Nuclease plus an end-protected dsODN tag in cells, then sequencing of tagged junctions | Amplicons you already decided to order |
| What it can find | Sequence lookalikes the model can see | Breaks that accepted the tag, including sites Tsai's then-available predictors missed | Whether a chosen coordinate is edited |
| Meaning of a quiet result | No lookalike passed the search settings — not a negative GUIDE-seq experiment | No tagged break recovered under those conditions — still not a complete genome, and not a clinical certificate | That amplicon looks quiet — a site you did not amplify was never asked |
When a Lab Reaches for GUIDE-seq
A predictor remains useful for nominating lookalikes and discarding obvious bad guides. GUIDE-seq is the experimental next step when the project's risk requires a cell-based, genome-wide look at actual breaks. A clean computational report is not that look. Tsai's unpredicted-majority result is the reason: most recovered sites in that 13-nuclease, two-line set were not on the lists then in use.
Not every dish knockout needs the assay. The trigger is risk plus the fact that prediction and capture answer different questions. GUIDE-seq does not replace the word "off-target." It measures tagged breaks after that word is already defined.
What GUIDE-seq Does Not Prove
A GUIDE-seq catalog is not a clinical-safety result. Tsai wrote that the method could be used to evaluate nucleases before clinical use. That is the authors' motivation, not a certificate this page adopts. The 2021 protocol says the method has been adopted for research and therapeutic use; that sentence is also not clearance. Biochemical assays can still find sites the tag assay misses.
Scoring software cannot run GUIDE-seq. ZettaCRISPR is one workspace example of on- and off-target scoring, and only that. The Zettalab product page documents CRISPR guide-RNA design with on- and off-target scoring. It does not document dsODN capture. If the remaining decision is which scoring environment to sit in, the CRISPOR versus commercial CRISPR software page compares those tools. None of them is this assay.
Frequently Asked Questions
What does GUIDE-seq measure that computational prediction misses?
Breaks that were tagged in cells, including sites the computational methods then in use did not list. Prediction lists lookalikes. GUIDE-seq reports recovered tags. Tsai's majority-missed result is scoped to 13 RNA-guided nucleases in two human cell lines.
How does GUIDE-seq capture CRISPR-induced breaks?
An end-protected double-stranded oligodeoxynucleotide is co-delivered so it can integrate at nuclease-induced double-strand breaks. Tag-containing DNA is then amplified and sequenced. That is a wet capture, not a score.
Does a clean GUIDE-seq run prove a guide is clinically safe?
No. GUIDE-seq is a cell-based catalog of tagged breaks under those conditions. The 2021 protocol notes that complementary biochemical assays can still find sites it misses. That is not a clinical-safety result.
Can CRISPR design software perform GUIDE-seq?
No. Design tools rank nominated guides. GUIDE-seq is a wet, cell-based assay. ZettaCRISPR is documented for on- and off-target scoring, not for running GUIDE-seq.
Related Articles