Sanger + TIDE/ICE: quantifying CRISPR editing without NGS

This is Arc 2, Part 12 of the CRISPR from Bench to Analysis series.

You ran your T7E1. You got bands. You know CRISPR cut something. Now you need a number — a real editing percentage you can put in your paper, report to your PI, or use to decide whether to expand that clone.

That's where TIDE and ICE come in.

Both tools take the same input: a Sanger sequencing trace from your edited sample and an unedited control. They analyze where the two traces diverge — right at the cut site — and mathematically decompose the mixed signal into an editing percentage. No NGS, no library prep, no waiting two weeks for a sequencing run. Just a PCR, two Sanger reactions, and an upload.

What you'll learn

• Why Sanger sequencing gives you a "mixed" trace after CRISPR editing
• How TIDE and ICE deconvolve that mixed signal into a real editing percentage
• Step-by-step: from PCR to result
• How to interpret your ICE score, R², and indel distribution
• Where TIDE/ICE falls short (and when you need Amplicon NGS instead)

The core idea: Sanger + deconvolution

After you transfect your CRISPR reagents and let cells repair for 48–72 hours, your cell population is a mixture. Some cells have the wild-type sequence. Others carry indels — insertions and deletions of various sizes and at slightly varying positions around the cut site.

When you PCR-amplify the target region and Sanger sequence the product, you get a trace that looks clean right up until the cut site. Then everything goes haywire: multiple overlapping peaks, one on top of the other, because the sequencer is reading a pool of sequences that diverge at exactly that point.

Before TIDE and ICE, that messy trace was nearly uninterpretable. You could see editing happened, but you couldn't quantify it. The insight behind both tools is that this mixed trace is actually informative — it's a weighted sum of the individual sequences in your pool. If you know what the unedited sequence looks like, you can subtract it out and figure out what's left.

That subtraction — comparing edited trace to unedited control, finding the divergence point, and using regression to fit the best combination of indel sequences to explain the leftover signal — is deconvolution.

TIDE

TIDE (Tracking of Indels by Decomposition) was developed at the Netherlands Cancer Institute and published in 2014. It was one of the first tools to make Sanger-based editing quantification practical.

Here is how the algorithm works at a conceptual level. You give TIDE two .ab1 trace files: your edited sample and an unedited control from the same primer set. TIDE aligns both traces up to the expected cut site. Before the cut site, the traces should be nearly identical — this region serves as an internal quality check. After the cut site, the edited trace becomes a mixture of the control sequence plus contributions from cells carrying +1, +2, −1, −2, −3, and larger indels.

TIDE uses linear decomposition (think linear regression) to find the combination of shifted and gapped sequences that, when summed together, best explains your edited trace. The result: a percentage assigned to each indel size, plus an overall % indel efficiency.

What TIDE outputs:

• Overall % indel (the number most people report)
• An indel spectrum: how much of the editing is +1 vs. −1 vs. −3 vs. larger deletions
• A decomposition quality score (R²) — how well the model fits your trace

TIDE is available at tide.nki.nl. It is free and requires no account.

ICE

ICE (Inference of CRISPR Edits) is Synthego's free tool and has largely become the community standard for Sanger-based CRISPR quantification. It uses the same conceptual approach as TIDE — deconvolving a mixed trace against an unedited control — but with a few practical advantages.

ICE also accepts two .ab1 files. You additionally supply the guide RNA sequence so it can pinpoint the expected cut site automatically (3 bp upstream of the PAM for SpCas9). This removes one source of user error.

What ICE outputs:

• ICE score: A 0–100 number representing the percentage of sequences with any detectable edit. This is the headline number for most experiments.
• R²: How well the deconvolution model fits the actual trace. A high R² (≥ 0.9) means the tool has a clean, reliable read. A low R² (< 0.8) is a warning sign.
• Indel distribution: A bar chart showing +1, −1, −2, etc., with their individual frequencies.
• HDR estimation (if you provide a donor sequence): ICE can estimate the fraction of sequences incorporating your HDR template, though this is an approximation.

ICE is available at ice.synthego.com. Free, no NGS required, results in minutes.

Step-by-step protocol

Step 1: PCR-amplify the target region

Design primers that give you a 300–800 bp product. The key constraint: your sequencing primer should sit roughly 150–250 bp away from the expected cut site. Too close and Sanger sequencing doesn't have enough run-up to resolve the trace before the cut site. Too far and the signal gets noisy. Most labs reuse the same primers they used for their T7E1 gel — if those worked, they will work here.

Run a clean PCR and check it on a gel. You want a single, tight band. Gel-extract or column-purify to remove primer dimers.

Step 2: Sanger sequence the PCR product

Submit two reactions to your sequencing provider:

1. Your edited sample
2. An unedited control — ideally the same cell type, same passage, same everything except no CRISPR reagents

Use the same sequencing primer for both. The control must come from your actual cells, not a plasmid or synthetic template — you want any natural SNPs in your cell line to be present in both traces so they cancel out.

Step 3: Download the .ab1 files

Your sequencing provider will return .ab1 (chromatogram) files. These are the raw trace files. Make sure you download the .ab1, not just a FASTA or text file — the tools need the actual peak intensity data.

Step 4: Upload to TIDE or ICE

For TIDE (tide.nki.nl):

• Upload your edited sample .ab1 as the "mutant" file
• Upload your unedited control .ab1 as the "control" file
• Set the decomposition window — TIDE will suggest a range, but make sure it brackets the cut site
• Set the indel range (default −10 to +10 is usually fine)
• Click Analyze

For ICE (ice.synthego.com):

• Upload your edited sample .ab1
• Upload your unedited control .ab1
• Paste your guide RNA sequence (20 nt, no PAM)
• If you used an HDR donor, upload the donor sequence (optional)
• Click Analyze

Step 5: Interpret the output

For a typical knockout experiment, you are looking for an ICE score ≥ 20% to confirm editing. For a clonal selection, you want to see whether your clone is homozygous (one sequence dominates) or heterozygous (two indels both present). The indel distribution tells you which cuts your guide favors — useful for understanding whether you are getting the frameshift you intended.

TIDE vs ICE: side by side

For most experiments, ICE is the tool to reach for first. It is more user-friendly, the guide sequence input lets it auto-detect the cut site, and the R² output gives you built-in quality control. TIDE is still useful when you want to do a quick check without needing to recall your guide sequence, or for batch analysis via the command-line version.

Interpreting your results

What is a good ICE score? In a bulk-transfected population, ICE scores of 20–60% are common. Above 60% is excellent guide efficiency. Below 10% — check your transfection, verify your guide sequence, and consider whether you are looking at the right amplicon.

What does R² below 0.9 mean? Low R² means the tool cannot cleanly fit the trace. Common causes:

• Sequencing failed or the trace is noisy (low signal, secondary peaks everywhere)
• Your editing efficiency is extremely high (>80–90%) — the "mixed" signal is no longer mixed enough for the algorithm to work well
• Your sequencing primer is too close to the cut site and the trace quality degrades right where the analysis happens
• The unedited control is not actually clean — it was also transfected

If your R² is below 0.8, re-sequence before trusting the result.

What does the indel spectrum tell you? The most common outcome with SpCas9 is a +1 insertion, followed by −1 and −2 deletions. If you see a lot of large deletions (−10, −20), your guide may be cutting at a region with unusual repair characteristics. The indel spectrum is also useful for predicting whether you are generating frameshifts: a +1 or −2 both cause a frameshift, while a −3 or −6 might leave the reading frame intact.

Limitations: where TIDE/ICE falls short

TIDE and ICE are excellent for screening experiments and for characterizing a single well-defined cut site. But there are real limits:

High editing efficiency (>80%): The deconvolution algorithm works by finding the difference between edited and unedited traces. When nearly every allele is edited, there are few unedited sequences left as a reference, and the subtraction gets unreliable. R² will drop and your estimate will be noisy.

HDR quantification: ICE can estimate HDR rates, but it is not designed for precision HDR measurements. If HDR efficiency is your primary readout, Amplicon NGS is more reliable.

Multiple alleles and complex edits: In diploid cells, you may have two different indels — one on each allele. TIDE and ICE can handle this reasonably well at the population level, but if you need to know exactly what happened on each chromosome in each clone, Sanger sequencing a pool is not sufficient.

Mosaic edits: If you edited post-zygotically (in vivo or during early development), you may have many different allele populations in the sample. The deconvolution model can struggle with >3–4 distinct sequences.

For any of these situations, Amplicon NGS is the right tool — which we will cover in CRISPR-13.

My Take

TIDE and ICE fill exactly the gap between a T7E1 gel and full Amplicon NGS. They take what was an unreadable Sanger trace and turn it into a publishable number. For 80% of CRISPR experiments — screening guides, confirming knockout in a cell line, checking a small cohort of clones — ICE is all you need.

The workflow is genuinely fast. If you already have PCR primers for your T7E1 gel, you can have an ICE score back within 24 hours of submitting your Sanger reactions. That kind of turnaround changes how you iterate on guide optimization.

Use it. Just know the R² score and don't trust results below 0.8 without re-sequencing.

Uploaded your trace to ICE and getting a low R² score? Describe your setup below and let's figure out what went wrong.

Resources