Most stories about CRISPR therapeutics make a quiet, flattering assumption. They celebrate the molecular scissors — the elegance of a guide RNA finding its target, the precision of a single base rewritten — and then stop before the harder question: once you can edit almost anything, how do you decide what is actually worth correcting?

That omission matters more than it seems. A genome can be sequenced in a day, a variant can be flagged in an afternoon, and a base editor can be designed in software by the weekend. None of that tells you whether the variant you found is the reason your patient is sick, whether correcting it is safe, or whether the edit will help. In practice, clinical interpretation — not editing — is the slow, decisive step between a sequence and a cure.

Table of Contents

• The Edit Was Never the Hard Part
  • Why the bottleneck moved
  • Why this matters at the bedside
• What Clinical Interpretation Actually Means
  • The five-tier verdict
  • The uncomfortable size of the gray zone
  • Why a variant can sit unresolved for years
• From Variant to Target: The Editability Question
  • Causal is not the same as correctable
  • How an approved therapy sidestepped the broken gene
  • The guide RNA is an interpretation, too
• The N-of-1 Sprint: Sequence to Cure in Months
  • What made the case fast
  • What the speed quietly assumed
• Why Faster Cures Raise the Stakes on Interpretation
  • Manufacturing outran curation
  • The reclassification problem nobody owns
  • A framework instead of false confidence
• The Deeper Meaning of Reading a Genome
  • Sequence is evidence, not destiny
  • Interpretation is a human act

The Edit Was Never the Hard Part

For most of the gene-editing era, the limiting reagent was technical capability. Could we cut DNA at a chosen site? Could we do it without shredding the rest of the genome? Those questions defined a decade of work. They are, for a growing list of conditions, now answered.

The first CRISPR-based therapy reached patients on that strength. Casgevy (exagamglogene autotemcel) was authorized in the United Kingdom in November 2023, then by the US FDA in December 2023 for sickle cell disease and in January 2024 for transfusion-dependent beta-thalassemia, with European approval following shortly after. It works by editing a patient's own blood stem cells at a regulatory region of the BCL11A gene, switching fetal hemoglobin back on. The editing itself is no longer the frontier. The frontier is everything that has to be decided before the scissors are allowed near the genome.

That is the shift this article is about. When editing was hard, interpretation could hide behind it. Now that editing is increasingly routine, interpretation stands exposed as the real rate-limiting step.

Why the bottleneck moved

Sequencing collapsed in cost and time. Editing tools matured. What did not scale at the same pace was our ability to say, with confidence, what a given sequence change means for a given person.

Consider the raw arithmetic. Of the roughly four million missense variants catalogued in humans, only a small fraction — on the order of a couple of percent — carry a clear clinical classification as either benign or pathogenic. The overwhelming majority sit in an undecided middle. Sequencing more genomes does not shrink that middle on its own; it usually enlarges it, because every new genome contributes private variants nobody has seen before.

Practical rule: A variant being editable tells you nothing about whether it is worth editing. Capability and indication are separate questions, and only one of them is solved by better tools.

Why this matters at the bedside

A clinician staring at a sequencing report is not asking an engineering question. They are asking a clinical one: does this change explain the disease in front of me, and would acting on it help or harm? CRISPR raises the temperature of that question, because for the first time the answer can translate directly into a permanent rewrite of the patient's DNA.

When the only response to a variant was watchful waiting, an ambiguous result was uncomfortable but rarely catastrophic. When the response can be a one-time, irreversible edit, the cost of misreading the sequence rises sharply. The tool got sharper. So did the consequences of pointing it at the wrong target.

If you want a broader place to reason through questions like that with other students and researchers, DNAnswer's molecular biology community is built around exactly these conversations.

What Clinical Interpretation Actually Means

Interpretation sounds soft next to the hard chemistry of editing, but it is a structured, evidence-weighing discipline. The field's shared grammar comes from the 2015 framework published by the American College of Medical Genetics and Genomics and the Association for Molecular Pathology — the ACMG/AMP guidelines — which laid out a standardized way to weigh evidence about a variant and reach a verdict.

The five-tier verdict

The framework sorts variants into five categories, and the labels do real clinical work:

Classification	Plain meaning	Clinical consequence
Pathogenic	Strong evidence it causes disease	Treated as a positive, actionable result
Likely pathogenic	Probable cause, not yet certain	Usually actionable, with caution
Variant of uncertain significance (VUS)	Evidence is insufficient or conflicting	Inconclusive — explicitly not actionable
Likely benign	Probably harmless	Treated as a negative result
Benign	Strong evidence it is harmless	Negative result

The two ends of that table are where medicine happens. The middle row — the VUS — is where medicine stalls. A VUS is not a mild finding. It is a formal statement that the evidence does not yet justify a decision, which means it should not drive treatment or even cascade testing of relatives.

The uncomfortable size of the gray zone

Here is the part most overviews soften. In some adult genetics practices, roughly half of all reported variants are of uncertain significance. For certain genes the picture is starker still: in one large analysis of the APC gene, around two-thirds of variants in a public database were VUS or carried conflicting interpretations.

That gray zone is not a rounding error. It is the central obstacle between sequencing and action. A VUS can also do quiet harm: studies of BRCA1 and BRCA2 testing have found that patients carrying an uncertain variant were sometimes managed more aggressively than those without one — even though many of those variants were later reclassified as benign.

A "variant of uncertain significance" is the most honest result in genomics and the most operationally frustrating. It tells you the truth: we do not yet know enough to act.

Why a variant can sit unresolved for years

To narrow the gray zone, the Clinical Genome Resource (ClinGen) convenes Variant Curation Expert Panels — gene- and disease-specific groups that adapt the general ACMG/AMP rules into sharper, context-aware criteria. When those refined criteria are applied, large batches of VUS can be reclassified into something actionable. One reassessment of tumor-suppressor variants reclassified roughly a third of stubborn VUS once newer, phenotype-aware criteria were applied.

But that work is expert-driven and slow. Panels cannot curate every variant, so they prioritize: conflicts in public databases, variants repeatedly submitted as uncertain, and variants likely to resolve cleanly given the available evidence. A private variant seen in a single family may wait years for the evidence that tips it one way or the other. Interpretation, in other words, is not a one-time calculation. It is an ongoing argument that the literature slowly settles.

From Variant to Target: The Editability Question

Suppose interpretation succeeds and you have a confidently pathogenic variant. You still are not done. A second layer of judgment sits between a confirmed cause and a viable edit, and it is easy to skip past.

Causal is not the same as correctable

A variant can be unambiguously disease-causing and still be a poor editing target. The gene might be too large to deliver. The relevant cells might be unreachable. Correcting the variant might require an edit that current tools cannot make cleanly. Causality is a claim about biology; correctability is a claim about what your delivery vehicle and editor can actually achieve in the right tissue.

• Causal answers: does this change drive the disease?
• Editable answers: can a tool reach the right cells and make the right change?
• Worth editing answers: does the benefit outweigh the risk of an off-target or incomplete edit?

Only when all three line up does a sequence become a credible therapeutic target.

How an approved therapy sidestepped the broken gene

Casgevy is the instructive example, because it does not correct the mutations that cause sickle cell disease. Instead of repairing the faulty hemoglobin gene, it disrupts a regulatory region of BCL11A in blood stem cells, reactivating fetal hemoglobin that compensates for the defective adult form. The interpretive insight was not "here is the broken letter" but "here is a switch we can flip to make the broken letter matter less."

That is a profound reframing. The most powerful editing strategy sometimes ignores the causal variant entirely and targets the biology around it. Deciding that is an act of interpretation, not chemistry.

Lab instinct: The best target is not always the broken gene. It is the most editable point in the pathway where a change does the most good with the least risk.

The guide RNA is an interpretation, too

Even the guide RNA — the sequence that aims the editor — encodes a set of judgments. Choosing where to cut or which base to convert involves predicting on-target activity and, just as importantly, anticipating off-target sites where the editor might act by mistake. A guide that performs beautifully in software can still find an unintended near-match elsewhere in the genome.

So the design step is not mechanical transcription of a target into a tool. It is a risk assessment expressed in nucleotides. Every guide is a hypothesis that this sequence, in this genome, will do this and only this.

The N-of-1 Sprint: Sequence to Cure in Months

The most vivid demonstration of the whole pipeline came in 2025, with a child known as KJ. Diagnosed within days of birth with severe CPS1 deficiency — a urea-cycle disorder in which ammonia builds to dangerous levels — he became, in February 2025, the first person in the world to receive a fully personalized CRISPR base-editing therapy, developed by teams at Children's Hospital of Philadelphia and Penn Medicine and reported in The New England Journal of Medicine.

The therapy was bespoke. After his specific CPS1 variant was identified, the team designed, manufactured, and delivered a base editor — carried to the liver by lipid nanoparticles — built to correct his particular mutation, with the first dose given between six and seven months of age and follow-up doses in the weeks after. The reported timeline from variant identification to a manufactured therapy was roughly six months.

What made the case fast

Several conditions had to be true at once, and they are worth naming because they are the real accelerators:

• The causal variant was identified quickly and confidently, removing the interpretive ambiguity that stalls most cases.
• The team had spent years building a platform for liver-targeted base editing, so the bespoke design slotted into existing groundwork rather than starting from zero.
• The disease mechanism was well understood, so correcting the variant had a clear, predictable benefit.

In other words, the editing was fast because the interpretation was already largely done. The sprint was possible only on a foundation of prior, slow, careful work.

What the speed quietly assumed

The headline was the months-long turnaround. The unstated premise was that the variant's meaning was certain enough to bet a child's genome on. Strip away that certainty and the whole timeline collapses, because you cannot responsibly manufacture a permanent edit against a variant you cannot confidently call pathogenic.

The remarkable thing about a six-month cure is not the editing. It is that the interpretation was solid enough to make the editing safe to attempt.

This is the lesson hiding inside the triumph. Acceleration did not come from skipping interpretation. It came from a setting where interpretation happened to be unusually clean.

Why Faster Cures Raise the Stakes on Interpretation

It would be comforting to think that as editing gets faster, the rest of the pipeline relaxes. The opposite is true. Speed on the manufacturing side increases the pressure on the interpretive side, because the gap between "we can build this edit" and "we are sure we should" becomes the riskiest part of the process.

Manufacturing outran curation

Designing and producing a personalized editor can now happen faster than the evidence base for a rare variant accumulates. That mismatch is new. For most of medicine's history, the slow step was treatment; now, for editable conditions, the slow step is being sure the target is right. When the build is fast and the verdict is slow, the temptation is to let confidence in the tool stand in for confidence in the target.

The reclassification problem nobody owns

Variant interpretation is not static. A variant called uncertain today may be reclassified as benign or pathogenic next year as new data arrive. The trouble is getting that update back to the patient. Analyses of clinical practice have documented cases where a variant's classification was meaningfully revised in a public database but the change was never communicated to the patient or their provider.

ACMG describes recontact and reinterpretation as a shared responsibility across labs, clinicians, and databases. In practice, shared responsibility drifts toward no responsibility, and updates fall through the cracks. In an editing era, that drift is dangerous: a permanent decision may rest on a classification that has since changed.

A framework instead of false confidence

When a sequence is proposed as a therapeutic target, it helps to interrogate it deliberately rather than trust the tooling.

Question	What it tests	Why it matters before editing
Is the variant confidently classified?	Strength of the pathogenic call	A VUS is not a valid editing target
Is the classification current?	Recency of the evidence	Old calls may have been quietly revised
Is the gene the best target?	Pathway-level strategy	The causal gene is not always the editable one
Does the guide act only where intended?	Off-target risk	A clean design can still find unintended matches
Does the benefit outweigh permanence?	Risk of an irreversible edit	Editing forecloses the option to reconsider

None of these are answered by the editor. They are answered by interpretation — the part of the pipeline that does not photograph well but decides everything.

If you want practice reasoning through evidence-weighing problems like these rather than memorizing definitions, DNAnswer's daily quiz on molecular biology concepts is a useful way to build the habit.

The Deeper Meaning of Reading a Genome

What makes this era genuinely new is not that we can edit DNA. It is that the distance between reading a genome and acting on it has shrunk to the point where interpretation can no longer hide behind technical difficulty. The judgment is now the bottleneck, in plain view.

Sequence is evidence, not destiny

A sequence is not a diagnosis. It is evidence that has to be weighed against population frequency, functional data, family history, and mechanism before it earns a verdict. The five-tier classification system exists precisely because a letter in the genome does not announce its own meaning. Most letters, in fact, refuse to.

The most important number in a genetic report is not how many variants were found. It is how many were understood well enough to act on.

That reframing matters for everyone downstream. A pathogenic call invites action. A VUS demands restraint. Confusing the two — treating uncertainty as if it were knowledge — is the failure mode that faster editing makes more costly, not less.

Interpretation is a human act

There is something quietly human in all of this. We built machines to read the genome, and then better machines to rewrite it. But the decision about what to rewrite still rests on argument, evidence, and the willingness to say "we do not yet know" when that is the honest answer. The baby treated in months was saved not by the speed of the editor but by the certainty of the interpretation behind it.

The era of CRISPR therapeutics will not be defined by how fast we can cut or convert DNA. That problem is increasingly solved. It will be defined by how honestly and how quickly we can move a variant out of the gray zone and into a verdict we are willing to act on. The question is no longer whether we can edit a sequence. It is whether we have understood it well enough to deserve to.