Skip to content
rnaprog
This methodology article is published in English; the interface and SEO metadata are localised. Read the original English version.

sgRNA design: PAM choice, spacer composition and the poly(T) trap

Why a spacer that scores well can still fail in cells, how PAM availability changes with your edit, and the U6 termination rule that catches people after ordering.

rnaprog editorial 7 min read

sgRNA design looks like a search problem: find a PAM, take the 20 nt upstream, check GC. The search is easy. The reasons guides fail are elsewhere.

PAM choice changes the answer

SpCas9 uses NGG. That is the default and the one with the most validated guides. But if your edit sits in an AT-rich region with no NGG nearby, you have options:

  • SpCas9-VQR (NGAN) or EQR (NGAG) variants relax the PAM at some cost to fidelity.
  • SaCas9 (NNGRRT) has a longer PAM, which reduces the candidate space but gives a smaller enzyme — useful for AAV delivery.
  • Cas12a (TTTV) cuts with a 5′ PAM, which is convenient for certain orientations and gives staggered cuts.

The point: the PAM is a design parameter, not a constant. If a target region returns no candidates at NGG, the answer is usually to change the enzyme, not to accept a poor guide.

What actually predicts guide activity

The Doench-style position rules are the best-documented heuristics: a G at position 20 of the spacer (adjacent to the PAM) helps, a C there hurts, and GC between 40% and 70% is the usable window. These are statistical associations from large screens, not mechanisms. They will rank guides correctly more often than chance and will still be wrong sometimes.

Two composition issues deserve more weight than the score:

Poly(T) of four or more. If your sgRNA is expressed from a U6 promoter, RNA polymerase III terminates at a run of four T’s. A spacer containing TTTT will produce a truncated guide that does not work — and nothing in a GC or on-target score will warn you. This is a hard filter, not a preference.

G-rich stretches at the 5′ end. GGG at the start of the spacer can impair transcription initiation and folding. Not fatal, but worth avoiding when an alternative exists.

Off-targets: what a design tool can and cannot say

A design run cannot tell you how many genomic off-targets a guide has. That requires aligning the spacer plus PAM against a reference genome and scoring mismatch positions — minutes of compute against a multi-gigabyte index. Any tool that reports a specific off-target count for a genome without doing that alignment is reporting something else (usually an internal repeat count).

What we do: rank by spacer composition, flag poly(T) and G-rich starts, and tell you plainly that a whole-genome search is out of scope. If your edit is going into a therapeutic programme, run the shortlist through an aligner or a dedicated off-target tool before ordering. It is the step that separates a research guide from a candidate.

Cut position matters for the edit

Cas9 cuts ~3 bp upstream of the PAM. If you are doing a knock-in by homology-directed repair, the intended edit should sit within ~10 bp of the cut site; if it is 40 bp away, repair efficiency collapses. Design the guide around the edit, not around the region.

For a knockout, the useful guides sit in an early constitutive exon, ideally within the first third of the CDS, and target a region shared by all transcript isoforms. A guide in an alternatively spliced exon will knock out one isoform and leave the others.

A short checklist before ordering

  1. Spacer is 20 nt immediately 5′ of a confirmed PAM in your target sequence — verify the PAM in the actual sequence, not from memory.
  2. GC 40–70%, no poly(T) ≥ 4, no GGG at the 5′ end.
  3. Cut site within 10 bp of the intended edit (HDR) or in an early shared exon (KO).
  4. Shortlist of 2–3 guides run through a genomic off-target search.
  5. If using a U6 cassette, confirm the spacer does not create a termination signal when combined with the scaffold junction.

Design is the cheap part. Ordering three guides instead of one costs a few hundred dollars; redoing a two-month experiment because the guide truncated costs considerably more.

La prima esecuzione parte con 300 crediti gratis

Basta per 6 costrutti mRNA o un pool di 12 siRNA, con QC ed export. Senza carta.

300 credits · No card required · You keep the rights to your sequences