Engine
Off-target analysis
The specificity check that follows guide design. Your shortlist is swept against the host genome with Cas-OFFinder, and each candidate comes back with a mismatch-count summary — enough to drop the weak options before anyone orders anything.
Cas-OFFinder
search engine
65
host genomes
0–4
mismatch tiers
Free
to run
01
Why the second pass exists
A guide that scores well on target has told you one thing and said nothing about the rest of the genome. Specificity is a separate question with a separate answer, and the cost of leaving it unanswered is paid much later, by someone else, in a growth room. This engine answers it while the shortlist is still a table.
It takes the candidates you starred in guide design and searches the host genome for near-matches using Cas-OFFinder. What comes back is not a verdict. It is a summary of where each candidate has company, counted by how many mismatches separate it from the sites it found, which is the form a reviewer can actually argue with.
02
What the report shows
The shortlist carries over
Candidates arrive from guide design already selected — no re-entry, no copy-paste between tools. You can also paste in a set from elsewhere, or point the sweep at every guide attached to a construct in the catalog.
Host genome and mismatch budget
Choose the host from the 65 supported organisms and set how many mismatches the search should tolerate, usually up to four. A wider budget returns more sites and takes longer; the interface tells you which you are asking for.
A mismatch-count summary per candidate
One row per candidate, with counts broken out by mismatch distance. A candidate with a single exact hit and nothing nearer than four mismatches reads very differently from one with several two-mismatch neighbours, and the table makes that difference obvious at a glance.
Context on every hit
Expanding a row lists the sites themselves with their coordinates, strand, and whether they fall inside an annotated feature. A near-match in an intergenic stretch is a different conversation from one sitting in a coding region.
Discard, keep, or replace
Candidates can be dropped from the shortlist directly in the report, and the gap sent back to guide design for a replacement. The shortlist that survives the sweep is the one the construct record inherits.
Dated and attached
Every sweep is stored against the construct version and the genome build it ran on. When an assembly is updated later, the platform can tell you which guides were cleared against the old build — which is exactly the kind of drift nobody notices unaided.
03
A worked example
- Shortlist
- 6 candidates, construct SS-0421
- Host genome
- Current build, from the organism library
- Mismatch budget
- Up to 4
- Candidates with near neighbours
- 2
- Discarded
- 2
- Carried forward
- 4
- Runtime
- A few minutes
- Engine class
- CPU — no credits
Four of the six came back clean enough to keep without discussion. The other two had company close enough to matter, and both were dropped rather than argued over — replacing a guide costs a click here and a great deal more anywhere else. Guide design was reopened, two replacements were starred, and the sweep was re-run on the amended shortlist.
That loop is the whole point. Specificity failures are cheap to find in a table and expensive to find in material, and the gap between those two costs is where this engine pays for itself.
04
How customers use it
Most teams run it as standard quality control: no guide enters a construct record without a sweep attached. Ag-biotech groups working through a series queue the sweeps as a batch and review the summary table once, which turns a per-guide chore into a single sitting. Groups with internal review requirements lean on the stored report, because it is dated, tied to a genome build, and travels with the export package when a design is handed to a lab.
The sweep also runs automatically inside design validation, so a project that has drifted past its last check gets flagged rather than assumed. Off-target analysis is a CPU engine and free to run at any volume. Credits — $25 for 250, one credit for roughly one Standard job — are for GPU engines elsewhere and are never spent here.
Ready to run it
Try this in the Splice Platform.
CPU engines are free to run. GPU engines run on prepaid credits.
Related