Genomic Data Is Not the Destination
Sequencing a tumour is now straightforward. Knowing what the result means for this patient, in this clinical situation, is the part that has not become easy.
4 August 2026 · 8 min read · Written for general
Fifteen years ago, sequencing a tumour was a research undertaking. Today a comprehensive panel can be run on a biopsy specimen as part of routine diagnostic work. The technical achievement is real and it has changed oncology.
It has also produced a characteristic new problem: a report arrives, it is dense with findings, and nobody is entirely sure which of them matters.
The shape of the problem
A comprehensive genomic profile typically returns a long list. Some entries are established drivers with a clear therapeutic association in that disease. Some are established drivers in a different disease, where the evidence does not transfer straightforwardly. Some are passenger alterations with no known consequence. A substantial number are variants of uncertain significance — detected, real, and of unknown meaning.
The report does not always make these categories obvious, and the visual weight given to a finding on a page does not reliably track its clinical weight.
Three questions the data does not answer by itself
Is this finding real? Every assay has a limit of detection, a coverage profile, and characteristic error modes. A variant at low allele fraction in a low-coverage region is a different kind of claim from a clean, high-fraction call. Tumour purity affects everything. So does specimen age and fixation.
Is this finding consequential? A gene being altered is not the same as the alteration having a functional effect, and a functional effect is not the same as a therapeutic implication. The step from "this gene is mutated" to "this drug may work" involves several inferences, each with its own strength.
Is this finding consequential here? This is the step most often skipped. BRAF V600E has a strong therapeutic association in melanoma. The same alteration in colorectal cancer behaves differently, and the treatment implications are not transferable in the way one might assume. Context is not a refinement on the molecular finding; it is part of what determines its meaning.
Evidence is uneven, and saying so is part of the job
Evidence tiering frameworks exist precisely because the strength of evidence behind molecular findings varies from randomised trial data to a single case report. A useful interpretation states which tier a finding sits in.
The temptation is to present all findings with equal confidence, because a report with hedging in it looks less authoritative. But the hedging is the information. A clinician told that a finding is supported by preclinical data alone can weigh it appropriately. A clinician told nothing about the evidence level cannot.
What the assay did not look for
A negative result means the assay did not detect something. Whether it means the thing is absent depends on whether the assay could have detected it.
Fusion detection is the clearest example. Several clinically important alterations in lung cancer and in sarcoma are gene fusions. An assay designed around sequence variants may not detect them reliably. A report that is silent on ALK is not the same as a report that excludes ALK, and the difference matters.
Copy-number changes, large structural rearrangements and certain repeat expansions have similar characteristics. "Not reported" and "not present" are different statements.
The destination
The point of sequencing a tumour is not to have sequenced it. It is to change a decision, or to establish with more confidence that a decision should not change.
That requires the data to be read in relation to the clinical situation, the evidence base, the treatment already given and the questions actually open. Sequencing is the beginning of that process. Treating it as the end produces a folder rather than an answer.