What do I need to run it?

Two Sanger traces from the same amplicon — one unedited control, one edited — plus the guide sequence. A donor or ssODN is optional and turns on the HDR readout.

Why do the allele percentages not add up to 100?

They sum to the model fit (R²). The unexplained variance is reported as unedited rather than distributed across alleles, which is what the original ICE does.

What is ICE-D, and why does it sometimes disagree with the ICE score?

ICE-D estimates editing from the discordant signal after the cut without fitting any alleles. It answers a different question from the regression, so a large gap between the two is itself a signal that the fit is poor.

Are my chromatograms uploaded anywhere?

No. The analysis runs locally in the browser, like everything else in the app.

Design

CRISPR Edit Inference

Work out what a CRISPR edit actually produced, from a pair of Sanger reads. Give it the unedited control trace, the edited trace and the guide, and it deconvolves the mixed chromatogram into the alleles behind it: the editing efficiency, each indel with its abundance and its sequence around the cut, the knockout score, and the HDR percentage when you supply a donor. The guide and its cut site are drawn on the construct map, and the per-base discordance profile shows exactly where the edited read separates from the control.

Open CRISPR Edit Inference Open the CRISPR edit inference (ICE) sample

A real edited amplicon and its wild-type control, decomposed into indel alleles

What it does

01Editing efficiency (ICE score) from a control and an edited Sanger trace
02Full indel spectrum: every fitted allele with its abundance and its sequence around the cut
03Knockout score — the share of alleles that frameshift or delete more than 20 bp
04HDR percentage when a donor or ssODN is supplied
05ICE-D discordance estimate, computed without the regression as an independent check
06Multiplexed edits: several guides at once, including two-cut dropout alleles
07Per-base discordance profile for both traces, with the alignment and inference windows marked
08Guide location and predicted cut site, on either strand, drawn on the construct map
09Allele table export as CSV

What it runs on

An independent implementation of the ICE method (Conant et al., Inference of CRISPR Edits from Sanger Trace Data, CRISPR J 2022) and of TIDE (Brinkman et al., Nucleic Acids Res 2014), checked against Synthego ICE 1.2.0 and the TIDE authors' reference code on every trace in its corpus, and against mixtures of known composition. Runs entirely in the browser — the traces are never uploaded. The panel runs on @ports/ice.

Everything runs client-side. Files are parsed in the browser and projects are saved on this machine, so an unpublished sequence is never uploaded.

Questions

What do I need to run it?: Two Sanger traces from the same amplicon — one unedited control, one edited — plus the guide sequence. A donor or ssODN is optional and turns on the HDR readout.
Why do the allele percentages not add up to 100?: They sum to the model fit (R²). The unexplained variance is reported as unedited rather than distributed across alleles, which is what the original ICE does.
What is ICE-D, and why does it sometimes disagree with the ICE score?: ICE-D estimates editing from the discordant signal after the cut without fitting any alleles. It answers a different question from the regression, so a large gap between the two is itself a signal that the fit is poor.
Are my chromatograms uploaded anywhere?: No. The analysis runs locally in the browser, like everything else in the app.

Also searched as

CRISPR indel analysis from Sanger · ICE analysis online · Synthego ICE alternative · TIDE alternative · CRISPR editing efficiency calculator · indel spectrum from chromatogram · knockout score from Sanger sequencing · HDR percentage from Sanger trace · ab1 CRISPR edit deconvolution · sequence trace decomposition CRISPR