CRISPR gRNA Design Rules
A good CRISPR guide RNA (gRNA) is a short spacer that sits next to a valid PAM, has moderate GC content, avoids sequences that break transcription or synthesis, and is unique in its target genome. The rules below match the transparent heuristics SeqBench's designer applies — use them to shortlist candidates before checking off-targets in a genome-aware tool.
| Parameter | Recommendation | Why |
|---|---|---|
| Spacer length | 20 nt (SpCas9); 21 nt (SaCas9); 23 nt (Cas12a) | Matches the nuclease's natural protospacer length |
| GC content | 40–70% ideal (30–80% usable) | Very low/high GC reduces activity and stability |
| Poly-T (TTTT) | Avoid | A run of ≥4 T's terminates Pol III (U6/H1) transcription of the guide |
| Homopolymer runs | Avoid runs ≥5 of any base | Hurt specificity and oligo synthesis quality |
| 5′ nucleotide | Prefer/append a G when using a U6 promoter | Pol III initiates efficiently from a 5′ G |
| PAM | Required, adjacent to the spacer (see PAM table) | No PAM, no cleavage; not part of the spacer you clone |
| Cut site | SpCas9: ~3 bp 5′ of the PAM (blunt); Cas12a: staggered, distal to PAM | Determines where the edit lands |
| Seed region | ~10–12 nt PAM-proximal — keep it unique | Mismatches here most strongly block off-target (and on-target) cutting |
| Off-target | Check genome-wide; minimize matches, especially in the seed | Specificity is genome-dependent and can't be judged from the guide alone |
Values are practical defaults — on-target efficiency and off-target risk vary by nuclease, locus and genome, so verify for your system.
Spacer length and PAM
Start by picking the nuclease, because it fixes the spacer length: 20 nt for SpCas9, 21 nt for SaCas9 and 23 nt for Cas12a. The spacer is the genomic sequence immediately adjacent to a valid PAM — see the CRISPR PAM table for each enzyme's PAM. Crucially, the PAM is not cloned into the guide: it must be present in the target DNA, but the spacer you order or express stops at the edge of the PAM.
Sequence rules (GC, poly-T, homopolymers, 5′ G)
Keep GC content in the 40–70% sweet spot (30–80% is usable); guides far outside this band tend to be less active or less stable. Avoid a poly-T runof four or more T's (TTTT): it is a Pol III terminator, so a U6- or H1-driven guide carrying it gets truncated during transcription. Likewise avoid homopolymer runs of five or more of any base, which hurt specificity and degrade oligo synthesis quality. When using a U6 promoter, prefer a guide that starts with G, or append a 5′ G, because Pol III initiates efficiently from a 5′ G.
Seed region and specificity
Not all positions in the spacer matter equally. The seed region — roughly the 10–12 nt closest to the PAM (PAM-proximal) — dominates target recognition. Mismatches in the seed most strongly block cutting, for both on-target and off-target sites, while mismatches in the PAM-distal end are better tolerated. A guide whose seed is unique in the genome is far less likely to drive off-target cleavage.
Choosing the target site
For a knockout, target an early, constitutive coding exon or a known functional domain so that an indel disrupts the protein in every transcript. Avoid the extreme 3′ end of the coding sequence, where a frameshift may leave most of the protein intact. Also avoid placing the spacer or PAM over a common SNP: a polymorphism in your samples can abolish targeting in some alleles or individuals.
Off-target checking
Off-target risk is genome-dependent and cannot be judged from the guide sequence on its own. Use the rules above to shortlist candidates, then verify each one in a genome-aware specificity tool before ordering oligos, paying special attention to near-matches in the seed region. Note that on-target efficiency models (for example Doench 2016) require a trained model rather than a simple heuristic, so treat any rule-based ranking as a first pass, not a final score.
Frequently asked questions
How long should a gRNA spacer be?
It depends on the nuclease: 20 nt for SpCas9, 21 nt for SaCas9 and 23 nt for Cas12a. Each value matches the natural protospacer length that nuclease recognizes. The PAM is adjacent to the spacer but is not part of the spacer you clone.
What GC content should a guide RNA have?
Aim for 40–70%, with 30–80% still usable. Very low or very high GC reduces guide activity and stability, so guides outside the usable band are best avoided when alternatives exist.
Why should I avoid TTTT in a guide?
A run of four or more T's (a poly-T stretch) acts as a Pol III terminator. When the guide is expressed from a U6 or H1 promoter, that signal cuts transcription short and you lose a functional guide.
What is the seed region of a gRNA?
The seed region is roughly the 10–12 nucleotides closest to the PAM (PAM-proximal). Mismatches here most strongly block cutting, so the seed dominates target recognition — keep it unique in the genome to limit off-target activity.
How do I check a guide for off-targets?
Off-target risk is genome-dependent and can't be judged from the guide sequence alone. Shortlist candidates using the rules here, then verify each one in a genome-aware specificity tool before ordering oligos.
Do I need a 5′ G on my guide?
When expressing the guide from a U6 promoter, a 5′ G helps Pol III initiate efficiently. If your spacer does not already start with G, you can prefer a guide that does or append an extra G at the 5′ end.
See also
Sources
- 1A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunityJinek M, Chylinski K, Fonfara I, Hauer M, Doudna JA, Charpentier E · Science 337(6096):816-821; doi:10.1126/science.1225829; PMID 22745249 · 2012Three table rows and the "Spacer length and PAM" section: the 20 nt SpCas9 spacer ("the 5'-terminal 20 nucleotides of the crRNA… are available for target DNA binding"); the Cut site row for SpCas9, "~3 bp 5' of the PAM (blunt)" ("cleavage produced blunt ends at a position three base pairs upstream of the PAM sequence"); and the PAM row — the PAM is in the target DNA adjacent to the protospacer ("one base pair downstream of the crRNA binding sequence"), which is what backs "not part of the spacer you clone" and FAQ 1 and FAQ 3. Also the origin of the seed concept: "a contiguous stretch of at least 13 base pairs between the crRNA and the target DNA site proximal to the PAM is required for efficient target cleavage."
- 2DNA targeting specificity of RNA-guided Cas9 nucleasesHsu PD, Scott DA, Weinstein JA, Ran FA, Konermann S, et al. (Zhang F) · Nature Biotechnology 31(9):827-832; doi:10.1038/nbt.2647; PMID 23873081 · 2013The Seed region row, the Off-target row, the "Seed region and specificity" section, FAQ 4 and FAQ 5. Verified quotes: "SpCas9 tolerates single-base mismatches in the PAM-distal region to a greater extent than in the PAM-proximal region" (backs "mismatches in the seed most strongly block cutting… mismatches in the PAM-distal end are better tolerated"); and on the seed's extent, "Single-base specificity generally ranges from 8 to 14 bp immediately upstream of the PAM, indicating a sequence-dependent, mismatch-sensitive boundary that varies in length." IMPORTANT NUANCE: this paper cites "a prototypical 10-12 bp PAM-proximal seed sequence" as the PRIOR model and its own data argue against a fixed length — so it backs "the seed dominates recognition" but qualifies the page's flat "~10-12 nt". It also backs the off-target claim empirically (>700 guide variants, >100 predicted off-target loci).
- 3Nucleotide sequences in Xenopus 5S DNA required for transcription terminationBogenhagen DF, Brown DD · Cell 24(1):261-270; doi:10.1016/0092-8674(81)90522-5; PMID 6263489 · 1981The Poly-T (TTTT) row's mechanism and the corresponding sentence in the "Sequence rules" section and FAQ 3 — specifically "a run of >=4 T's terminates Pol III (U6/H1) transcription". The abstract states it verbatim: "Termination occurs within clusters of four or more consecutive T residues in the noncoding DNA strand sequence." This is the primary origin of the four-T rule of thumb.
- 4Genome engineering using the CRISPR-Cas9 systemRan FA, Hsu PD, Wright J, Agarwala V, Scott DA, Zhang F · Nature Protocols 8(11):2281-2308; doi:10.1038/nprot.2013.143; PMID 24157548 · 2013The 5' nucleotide row, the corresponding sentence in "Sequence rules", and FAQ 6. Verified verbatim: "the U6 RNA polymerase III promoter used to express the sgRNA prefers a guanine (G) nucleotide as the first base of its transcript", and "an extra G is appended at the 5' of the sgRNA where the 20-nt guide sequence does not begin with G" — which is exactly the page's "prefer a guide that starts with G, or append a 5' G". Also independently backs the 20-nt guide length. This is a protocol paper, i.e. the primary source for the PRACTICE, which is what the page states.
- 5Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9Doench JG, Fusi N, Sullender M, Hegde M, Vaimberg EW, et al. (Root DE) · Nature Biotechnology 34(2):184-191; doi:10.1038/nbt.3437; PMID 26780180 · 2016Three things. (a) The "Doench 2016" named in the "Off-target checking" section as an on-target efficiency model requiring a trained model — this IS that paper (Rule Set 2), and it also supplies the CFD off-target metric ("the CFD score performed best… with AUC values ranging from 0.82-0.98"). (b) The Poly-T row as a DESIGN rule, independent of the 1981 mechanism: "we excluded sgRNAs with… a run of four or more thymidines." (c) Partially, the "Choosing the target site" section's advice to avoid the extreme 3' end of the CDS: "only the C-terminal 10% of the protein-coding region showed a statistically significant reduction in activity."