Is CBL Really a Melanoma Driver? Variant Interpretation Under Scrutiny in Stage II Disease

A quiet but consequential dispute has erupted in the pages of the British Journal of Cancer over one of the most consequential labels in modern oncology: the designation of a gene as a cancer driver. At stake is CBL, a gene proposed by a research team led by Lindner and colleagues as a novel driver and prognostic biomarker in stage II melanoma, the surgically resectable but clinically unpredictable intermediate stage of the disease. In a correspondence published in October 2026, two pathologists from Singapore, Kok-Siong Poon of the National University Hospital and Evelyn Siew-Chuan Koay of the Yong Loo Lin School of Medicine, argue that the evidence supporting CBL as a driver is far weaker than the original study suggests, and that the case exposes broader weaknesses in how cancer genomics interprets variants.

The original study made a striking claim. By profiling the genomic landscape of stage II melanoma, the researchers identified deletions at chromosomal region 11q23.1-q23.3 as a potential prognostic marker in RAS-mutant melanoma, and they reported an enrichment of small variants in CBL, particularly within the NF1 molecular subtype of the disease. Because CBL encodes an E3 ubiquitin ligase that negatively regulates receptor tyrosine kinase and MAPK signalling, the same pathway that drives melanoma through BRAF, NRAS and NF1 alterations, the gene is a plausible candidate. Losing a brake on this pathway could, in principle, fuel tumour growth. The study also reported a significant mutation hotspot at positions 417 and 418 in the RING finger domain of the protein, the structural element that gives the ubiquitin ligase its catalytic power, and robust in-silico predictions that these changes are damaging.

Poon and Koay do not dispute that CBL deserves attention. Their concern is that plausibility is not proof, and that the presentation of the variant data in the original paper leaves too many questions unanswered for the driver label to stand. Their critique begins with a deceptively simple technical point: the variants in the study’s main table were reported using only protein-level nomenclature, without the corresponding cDNA descriptions. Without cDNA notation, independent researchers cannot reliably verify the exact nucleotide changes, cross-reference databases, or reproduce the analysis. In a field where reproducibility is the currency of credibility, that omission matters.

The second concern cuts deeper. No variant allele frequencies were provided for the CBL mutations. Variant allele frequency, the fraction of sequencing reads carrying the mutant allele, is the primary tool for judging whether a variant is clonal, present in most tumour cells and therefore likely an early, cancer-relevant event, or subclonal and possibly a passenger. It is also the first line of defence against sequencing artefacts. This is especially important because the original study relied on formalin-fixed paraffin-embedded tissue, the standard archival material in pathology but notoriously prone to chemical damage that can generate false-positive mutations. Without allele frequency data, readers cannot distinguish a genuine tumour mutation from an artefact of fixation and sequencing chemistry.

The copy-number context of some variants raises further puzzles. Several CBL variants coexisted with heterozygous deletions of the region, a pattern consistent with a loss-of-function driver, since deleting the remaining copy of a tumour-suppressor-like gene is a classic two-hit mechanism. But one case showed coexistence with an amplification instead. For a hypothesised loss-of-function driver, amplification of the very locus supposed to be inactivated is biologically incongruous, and it forces the question of what CBL is actually doing in that tumour. The gene’s role may be context-dependent: in some settings CBL loss promotes signalling, while in others CBL has been implicated in oncogenic functions, which complicates any simple narrative.

Database evidence also came under scrutiny. The COSMIC catalogue counts cited in support of the variants were not consistently melanoma-specific, meaning mutations observed in entirely different tumour types were being marshalled as evidence of relevance in skin cancer. Similarly, the OncoKB classifications used to argue oncogenicity were based on evidence from other cancers. A variant that is oncogenic in myeloid leukaemia or another CBL-linked malignancy is not automatically oncogenic in melanoma, where cellular context, co-mutations and selective pressures differ profoundly. Cross-cancer evidence can generate hypotheses, Poon and Koay argue, but it cannot substitute for melanoma-specific proof.

Perhaps the most statistically pointed objection concerns recurrence, a cornerstone criterion for driver nomination. Genuine drivers are hit repeatedly by tumours because mutating them confers a selective growth advantage, the defining idea that Stratton, Campbell and Futreal articulated in their landmark 2009 Nature paper on the cancer genome. Yet of the 24 CBL variants found across 20 tumours in the study, 21 were unique. That near-total absence of the same variant appearing twice sits uneasily with driver status and suggests the enrichment signal may be driven by aggregation of heterogeneous, individually rare alterations rather than by repeated targeting of functionally critical residues. The correspondents also ask whether excluding variants of uncertain significance, the notorious VUS category, from the enrichment analysis would change the statistical significance of CBL as a driver. Including unclassified variants in such calculations can inflate or bias statistical signals, since VUS have not been shown to alter protein function at all.

Underlying the entire exchange is a structural gap in the variant-interpretation infrastructure. No ClinGen Gene Curation Expert Panel currently exists for CBL, the specialised bodies that convene to establish rigorous, gene-specific rules for judging pathogenicity. In their absence, interpretation of CBL variants must lean on general somatic variant guidelines, including the joint consensus recommendations of the Association for Molecular Pathology, the American Society of Clinical Oncology and the College of American Pathologists, and the more recent oncogenicity standards from ClinGen, the Cancer Genomics Consortium and the Variant Interpretation for Cancer Consortium. Those frameworks are powerful, but they are designed to be applied variant by variant with careful evidence weighing, not to rubber-stamp an entire gene as a driver on the strength of an enrichment statistic.

The stakes extend well beyond an academic quarrel. Driver gene designations shape prognostic stratification, therapeutic development and, eventually, clinical decision-making. If CBL deletions at 11q23.1-q23.3 genuinely mark aggressive RAS-mutant stage II melanomas, patients carrying them might be candidates for intensified surveillance or adjuvant therapy, and the CBL protein itself, as a regulator of RTK-MAPK signalling, could become a drug target. But if the signal is an artefact of FFPE sequencing, unverified nomenclature, cross-cancer evidence borrowing and VUS-inflated statistics, then building a biomarker on it risks misclassifying patients. The distinction between driver and passenger is not a stylistic preference; it is the foundational inference of cancer genomics, and it demands evidence that an alteration actually confers a selective growth advantage.

The path forward, as Poon and Koay make clear, runs through the laboratory rather than through larger statistical models. Functional studies are needed to validate the oncogenicity of the specific CBL variants, particularly the RING finger hotspot at positions 417 and 418, in melanoma-relevant cellular contexts. Variant-level interpretation should follow established somatic classification standards, with cDNA-level reporting, allele frequency transparency, melanoma-specific database comparisons and careful handling of copy-number co-occurrence. Until such work is done, the correspondents conclude, CBL should be regarded as a candidate requiring validation rather than an established driver in melanoma. Their letter is a reminder that in precision oncology, the hardest step is often not finding a candidate gene but proving that the tumour, and not the method, put the mutation there.

Subject of Research: Variant interpretation and driver gene designation for CBL in stage II melanoma

Article Title: CBL in stage II melanoma: considerations of variant interpretation for driver gene designation

Article References: Poon, K.-S., & Koay, E. S.-C. (2026). CBL in stage II melanoma: considerations of variant interpretation for driver gene designation. British Journal of Cancer. https://doi.org/10.1038/s41416-026-03629-1

Image Credits: AI Generated

DOI: 10.1038/s41416-026-03629-1

Keywords: CBL, stage II melanoma, driver genes, variant interpretation, cancer genomics, 11q23 deletions, RTK-MAPK signalling, VUS, FFPE artefacts, ClinGen, OncoKB, British Journal of Cancer