Discriminase — commensal-sparing CRISPR guide design

How to use

Set your nuclease (Parameters). Cas12a or SpCas9, or a custom PAM — this decides which sites a guide can target.

Build a panel of commensals to spare (Step 1). Add organisms on the right; your panel builds up on the left. Then press Build commensal index.

Add your target(s) to hit (Step 2). Search a name, add by accession/taxid, or paste/upload a sequence — add as many as you like. Press Find common target guides; with several targets you get only the guides that hit every one.

Press Run (Step 3). You get guides that hit the target(s) but avoid every commensal — download as CSV.

Ways to fill the panel

Load a panel → Examples — ready-made lists we ship (e.g. Gut microbiome, 52 species). Loads the organisms; you press Build. The genomes download in your browser the first time, then they're cached.

Load a panel → Prebuilt — a panel we already indexed for you: it's ready instantly, no Build step. (Appears here only when we've shipped one.)

Load a panel → Saved — panels you saved in this browser with “Save panel to browser”, so you can reuse them. Loads the list; you press Build.

Upload panel CSV — your own list, as a CSV with a taxid (or accession) column. Loads the list; you press Build.

One by one — search NCBI, or add a single accession / taxid, or upload FASTA genome files.

Saved vs Prebuilt, in one line: Saved = a list you kept and still need to Build. Prebuilt = already Built by us, instant.

💡 Off-target tolerance = how many mismatches still count as “would also cut a commensal.” Lower = stricter and safer. Nothing leaves your computer except genome downloads from NCBI.

What each parameter means

A guide sits next to a fixed PAM motif. The bases right beside the PAM are the seed — the stretch that matters most for whether the nuclease actually cuts. Below is a Cas12a site (PAM on the 5′ side):

5' TTTV ACGTACGTAC GTACGTACGTACG 3' ^^^^ ^^^^^^^^^^ ^^^^^^^^^^^^^ PAM seed rest of guide └──────── guide ───────┘

For SpCas9 the PAM sits on the other side (3′, after the guide); the seed is still the part of the guide nearest the PAM.

Nuclease preset — quick Cas12a / SpCas9 defaults, or Custom to set the PAM yourself.

PAM & PAM position — the fixed motif (e.g. TTTV) and which side of the guide it sits on.

Guide length — how long the whole guide block is (bases, PAM not counted).

Seed length — how many PAM-proximal bases count as the seed (the left part of the guide above).

Seed mismatch tolerance — mismatches allowed inside the seed before a commensal counts as “would be cut.” Usually strict (low).

Off-target tolerance — mismatches allowed across the whole guide for that same test.

Min / Max GC — allowed fraction of G+C bases in the guide (binding strength).

Max target guides — stop after screening this many of the target's guides (0 = all). Trades speed for completeness.

NCBI email — optional; identifies you to NCBI for faster, more reliable genome downloads.

⚙️ Parameters — Cas12a defaults (click to change nuclease / strictness)

Nuclease preset picks Cas12a / SpCas9 defaults (PAM + position), or Custom to set them yourself.

PAM (IUPAC ok)the fixed motif next to every guide. Need a spacer between PAM and guide? Pad it with N, e.g. TTTN.

PAM positionwhich side the fixed PAM motif sits on relative to the guide.

Guide lengthlength of the guide (protospacer) in bases, not counting the PAM.

Seed lengthhow many PAM-proximal bases count as the seed — the part most critical for cutting.

Seed mismatch tolerancemismatches allowed in the PAM-proximal seed — the part that matters most for cutting. Lower = stricter.

Off-target tolerance (total mismatches)total mismatches across the whole guide that still count as “would also cut a commensal.” Lower = stricter and safer.

Min GClowest allowed fraction of G+C bases — too low and the guide binds weakly.

Max GChighest allowed fraction of G+C bases — too high gets sticky and off-target-prone.

Max target guides (0 = all)stop after screening this many of the target's guides — keeps big genomes fast. 0 screens every guide.

NCBI email (optional)NCBI gives identified callers higher request limits, so genome searches and downloads go faster and fail less. Leave blank to stay anonymous.

1 Commensals to spare

The bacteria the guide must not cut. Add them on the right; your panel builds up on the left.

Your panel

Organism	Source	Status

Add commensals

Start from a whole panel

Load a panel

Upload panel CSV

…or add organisms one by one

Search NCBI by name

Accession

Taxid

Upload FASTA (one or more genomes)

2 Targets to hit

The bacteria you want to cut. Add one or more on the right; with several targets we keep only the guides that hit every one (exact match), then screen those against your panel.

Your targets

Organism	Source	Status

Add targets

…add organisms one by one

Search NCBI by name

Accession

Taxid

Upload FASTA (one or more genomes)

…or paste a DNA sequence

Target a region (optional — single target only)

Start (bp)

End (bp)

With one target, search only guides between these positions — e.g. inside one gene. Ignored when you have several targets (their coordinates differ). Tip: to target a specific gene precisely, paste/upload just that gene's sequence above.

3 Find commensal-sparing guides

Screens every target guide against the panel and keeps the ones that hit nothing in it.

Runs entirely in your browser. Only genome downloads go to NCBI. source & docs

Created by Owen Loh — website · LinkedIn · owenloh0607@gmail.com
Questions, ideas, or improvements? Get in touch.