CRISPR guide finder

Finds every 20 nt SpCas9 protospacer with an NGG PAM, on both strands.

Guide RNAs

How to find a CRISPR guide RNA

SpCas9 cuts wherever the 20 nucleotides of its guide RNA match the DNA and the three bases immediately after the match read NGG, the protospacer adjacent motif. Finding guides is therefore a search for NGG: every NGG with at least 20 bases in front of it gives one candidate.

5'-NNNNNNNNNNNNNNNNNNNN-NGG-3'  target DNA
    20 nt protospacer   PAM
                    ↑ blunt cut, 3 bp from the PAM

The guide RNA itself is the 20 nt protospacer with U in place of T, joined to the constant scaffold; the PAM is in the DNA only and is never ordered as part of the oligo. Both strands carry PAMs, so the reverse complement is searched as well and its guides are reported at their position on the plus strand of what you pasted.

Worked example

"Try an example" loads the 720 nt coding sequence of EGFP and finds 119 guides,47 on the plus strand and 72 on the minus strand. The first isAAGGGCGAGGAGCTGTTCAC at 10 to 29 with the PAM CGG, GC 60 %, cutting after base 26. GC across the set runs from 45 % to 80 % and no guide carries a warning, which is what a GC-rich coding sequence looks like. Turn "Search both strands" off and 47 are left.

What the flags mean

FlagWhat it meansWhat to do
TTTTA run of four or more T inside the protospacerAvoid for a U6 or H1 expressed guide; fine as synthetic RNA
GC lowUnder 30 % GCExpect weak binding and low activity
GC highOver 80 % GCExpect more off-target cutting
RepeatedThe same 20 nt occurs elsewhere in the inputIt will cut in more than one place, so pick another

Choosing between the guides

  • Position first. For a knockout, cut in an early constitutive exon, in the first third of the coding sequence, so that a frameshift removes the whole protein. For a knock-in, cut within about 10 bp of the insertion point, because repair efficiency falls off quickly with distance.
  • Then off-targets. The shortlist has to go through a genome-wide search. A guide with a perfect second site, or with several sites differing by one or two bases away from the PAM, is not usable however good it looks here.
  • Then the sequence itself. Skip the flagged guides, and prefer a G at the 5-prime end for a U6 promoter, which starts transcription on a G; otherwise one is usually added, giving a 21 nt guide.
  • Order two or three. Activity varies several fold between guides that look equally good, and no published score predicts it reliably.

Related tools

To check a guide or a repair oligo for self-structure and dimers, use theprimer dimer and hairpin checker. To search the target for any other motif, including a different PAM written in IUPAC codes, usepattern find. To design the primers that amplify the region for sequencing, use the Tm calculator andPCR product size. To see where a cut falls in the reading frame, use DNA to protein translation.

Frequently asked questions

What is a PAM and why does SpCas9 need one?

The protospacer adjacent motif is a short sequence in the DNA, immediately 3-prime of the target, that the Cas9 protein itself recognises. Streptococcus pyogenes Cas9 needs NGG. Without it the protein does not unwind the duplex, so no guide can be designed for a site with no NGG next to it. The PAM is in the DNA only, never in the guide RNA.

Where does SpCas9 cut?

Three base pairs from the PAM, between the 17th and 18th base of the 20 nt protospacer, and on both strands at the same place, leaving a blunt double-strand break. The tool gives that position on the plus strand for every guide, which is where an insertion or a deletion will be centred.

What GC content should a guide have?

Between about 40 and 70 % works well for most guides. Below 30 % the guide binds weakly and tends to be inactive; above 80 % it binds too well and off-target cutting rises. Both extremes are flagged here. GC alone is a weak predictor, so treat it as a filter that removes bad guides rather than a score that finds good ones.

Why are TTTT runs a problem?

A guide expressed from a U6 or H1 promoter is made by RNA polymerase III, which stops at a run of four or more T. A TTTT inside the protospacer truncates the transcript, so the guide is never made at full length. Guides delivered as synthetic RNA or as ribonucleoprotein are not affected, but the flag is worth heeding for any plasmid or lentiviral construct.

Does this check for off-target sites?

Only inside the sequence you paste: a guide whose 20 nt protospacer occurs more than once in the input is flagged as repeated. Real off-target prediction needs the whole genome and a mismatch model, so run the shortlist from here through a genome-aware tool such as CRISPOR or Cas-OFFinder before ordering anything.