DNA-Negative Driver Fusions: Adding Reflex RNA-Seq

The corresponding DNA-only figure was 4.8% for fusions detected exclusively through DNA profiling. The result is not a rejection of DNA-based next-generation sequencing; it is a demonstration that the two assays interrogate different biological layers and therefore produce different blind spots.

That distinction has become operationally important in precision oncology. A DNA panel may provide robust detection of single-nucleotide variants, small insertions and deletions, and many copy-number alterations, while still failing to resolve a therapeutically relevant rearrangement when its breakpoint lies in a large or poorly captured intronic region. Reflex RNA sequencing for gene fusions addresses that specific limitation by examining the expressed transcript rather than attempting to infer the event from a difficult genomic interval.

The intronic bottleneck: why DNA-NGS misses actionable fusions

A fusion is clinically useful only when the laboratory can identify the relevant rearrangement with sufficient analytical confidence and translate it into a treatment decision. The biological event may be straightforward: two genomic regions become joined, producing an abnormal transcript or activating a kinase domain. The technical problem is that the breakpoint does not necessarily occur in a compact, easily captured coding exon.

DNA-based NGS assays generally rely on hybrid-capture probes or amplicons positioned across selected genomic regions. For single-nucleotide variants, the assay can interrogate a defined sequence window and identify a nucleotide-level change. Fusion detection is more dependent on the architecture of the genes involved. If a breakpoint occurs within a large intron, the assay must either cover that intronic territory directly or use a design capable of detecting the rearrangement without requiring exhaustive coverage of every possible breakpoint.

Large introns create several failure modes:

• A capture design may not cover the full range of possible breakpoint locations because the non-coding region is too large for economical, uniform probe placement.
• Amplicon-based methods can fail when the breakpoint falls outside the primer configuration used by the panel.
• The rearranged DNA fragment may be difficult to interpret if the breakpoint is complex, repetitive, or accompanied by additional structural changes.
• Low tumor fraction, degraded formalin-fixed paraffin-embedded material, and uneven coverage can further reduce confidence in a borderline signal.
• A DNA assay may detect a rearrangement-associated copy-number pattern without establishing whether a functional, expressed fusion transcript is present.

RNA sequencing approaches the problem from a different direction. Rather than sequencing the entire intronic interval, targeted RNA methods evaluate the spliced coding exon junctions present in the expressed transcript. The assay therefore operates on the molecular product that may drive the cancer phenotype. This does not make RNA sequencing universally superior: RNA is more labile than DNA, expression is variable, and the assay remains dependent on tissue quality and transcript abundance. It does, however, make RNA particularly effective for many fusion classes in which the clinically relevant evidence is an expressed exon-to-exon junction.

The resulting paradigm is complementary rather than substitutive. DNA sequencing remains necessary for a broad range of genomic alterations, while RNA sequencing can determine whether a rearrangement produces a transcript with therapeutic relevance.

The central issue is not whether DNA or RNA is the better platform; it is whether the assay interrogates the molecular event at the level where it can actually be resolved.

This distinction also affects the interpretation of a negative result. A DNA-negative report does not necessarily mean that no actionable fusion exists. It may mean that the assay did not capture the breakpoint architecture, that the rearrangement was below the detection threshold, or that the relevant event was not detectable through the panel’s genomic design. The report should therefore be understood within the analytical scope of the assay, not as an absolute statement about the tumor’s fusion status.

Clinical impact of integrating RNA sequencing into precision oncology

The practical value of reflex RNA-seq is measured by incremental clinical yield: the number of actionable alterations found after adding RNA analysis to a DNA-first workflow, particularly when those alterations change the treatment landscape.

In the Tempus clinical dataset, 616 of 2,118 actionable level 1 and 2 fusions were detected exclusively through RNA sequencing. That corresponds to 29.1% of the actionable fusion events in the evaluated group. Only 4.8% were detected exclusively through DNA sequencing. The figures should not be converted into a universal performance claim for every laboratory because panel content, sample composition, sequencing chemistry, tumor types, and bioinformatics pipelines vary. They do establish the scale of the asymmetry: RNA can recover a clinically meaningful set of fusions that a DNA-only workflow may not report.

The clinical significance is concentrated in rearrangements involving actionable kinase or signaling pathways. Depending on tumor type and biomarker context, relevant fusion partners may involve genes such as ALK, ROS1, RET, NTRK1, NTRK3, FGFR2, BRAF, or NRG1. The therapeutic consequence is not merely classificatory. A detected fusion can support targeted therapy selection, alter the interpretation of a tumor’s driver landscape, and prevent the patient from being assigned to a less specific treatment strategy solely because the DNA assay was negative.

This is particularly consequential when the alternative pathway is characterized by treatment exhaustion. In advanced disease, a missed driver alteration can lead to a sequence of decisions based on incomplete molecular stratification. The laboratory cannot determine the clinical value of a fusion that it never identifies, and the oncologist cannot act on a biomarker absent from the report. Reflex testing reduces that information gap by making RNA analysis part of the diagnostic pathway rather than an exceptional request initiated after an unexplained treatment course.

Reflex does not mean indiscriminate

A reflex RNA strategy can be designed in several ways. The laboratory may trigger RNA sequencing for every specimen undergoing a DNA fusion-capable panel, or it may use tumor-specific rules based on histology, clinical indication, DNA findings, and the known limitations of the first-line assay. Each approach produces a different balance of tissue use, turnaround time, cost, and incremental yield.

A universal reflex pathway has a clear operational advantage: it reduces dependence on individual recognition of a possible fusion and minimizes the risk that the second assay is never ordered. A selective pathway may conserve tissue and laboratory capacity, particularly when the DNA panel already has strong fusion coverage for the relevant tumor type. The appropriate design is therefore an implementation decision, not a purely technological conclusion.

A useful reflex policy should define:

• Which tumor types receive automatic RNA testing after DNA-based profiling.
• Whether the RNA assay is performed in parallel or only after a DNA-negative or indeterminate result.
• Which fusion classes are considered clinically actionable within the laboratory’s reporting framework.
• How insufficient RNA quality, low expression, and ambiguous read support are handled.
• Whether the assay can distinguish an expressed in-frame fusion from a rearrangement with uncertain functional consequence.
• How positive findings are orthogonally confirmed when required by the laboratory’s validation plan.
• How the result is integrated into the final pathology and molecular report.

The phrase targeted RNA fusion panel reflex is therefore more precise than a generic instruction to add RNA sequencing. The panel must be designed around the fusion spectrum relevant to the laboratory’s case mix, and its bioinformatics pipeline must be validated for the structural patterns it claims to detect.

Tumor-specific yield: where RNA-seq outperforms DNA profiling

The incremental yield of RNA sequencing is not evenly distributed across all malignancies. Tumor biology and genome architecture determine how often clinically relevant fusions occur and how difficult they are to detect through DNA interrogation.

The available clinical data show particularly large RNA-only contributions in low-grade glioma and sarcoma. In low-grade glioma, 69.4% of fusions in the evaluated group were detected only through RNA sequencing. In sarcoma, the corresponding figure was 58.2%. These values illustrate why a uniform assumption about DNA panel performance is unsafe. A workflow optimized for common epithelial tumors may not provide the same fusion sensitivity in diseases where rearrangements are central to classification or where breakpoint architecture is unusually complex.

The glioma and sarcoma findings are not simply arguments for ordering more tests. They support tumor-specific assay governance. In a tumor type with a high probability of a diagnostically important rearrangement, RNA sequencing may function as a core component of the molecular workup. In another setting, it may be most efficient as a reflex assay triggered by a DNA result that is negative, incomplete, or inconsistent with the morphology and clinical context.

The same logic applies to non-small cell lung carcinoma. Lung cancer workflows are already structured around multiple actionable biomarkers, and failure to identify a targetable fusion can have direct therapeutic consequences. A combined DNA and RNA strategy increases the breadth of fusion detection, but it also introduces a measurable delay. In one evaluated dual-reflex workflow, the turnaround time increased from 8 working days to 10 working days.

Two working days can be clinically acceptable when they prevent a much larger diagnostic delay later. It can also be problematic when tissue is limited, the patient requires an immediate treatment decision, or the RNA assay is initiated without a clear reporting pathway. The value proposition depends on whether the incremental information is delivered within the decision window in which it can be used.

The turnaround-time tradeoff is a workflow problem

The addition of RNA sequencing is often described as though it were a single laboratory step. In practice, it changes specimen triage, nucleic-acid extraction, library preparation, quality control, bioinformatics review, variant interpretation, and report sign-out. The turnaround-time effect depends on whether DNA and RNA are extracted together, whether the assays are run in parallel, and whether the laboratory has a validated process for resolving discordant findings.

For a DNA-first reflex model, the sequence may look operationally simple: perform DNA NGS, assess whether fusion detection is adequate, and send selected cases to RNA sequencing. That model is tissue-efficient in some settings, but it delays RNA testing until the initial DNA result is reviewed. Parallel testing reduces the chance of a second delay but consumes more tissue and laboratory resources upfront.

The decision should be based on the clinical role of the assay rather than on a preference for one workflow architecture. A laboratory considering implementation should map the following points:

1. Specimen triage. Tissue must be allocated before extraction, with sufficient material reserved for RNA when the tumor type or clinical indication makes fusion detection consequential.

2. RNA quality control. The laboratory needs predefined thresholds for input quantity, integrity, library complexity, and usable transcript content. A failed RNA assay should not be reported as a biologically negative result.

3. Bioinformatics review. Fusion callers must distinguish plausible driver events from read-through transcription, recurrent technical artifacts, and low-support chimeric reads.

4. Orthogonal confirmation. The confirmation strategy should be defined during validation, particularly for rare fusions, novel partners, or findings with an immediate therapeutic implication.

5. Interpretive integration. The molecular report should reconcile DNA and RNA findings rather than presenting them as disconnected assay outputs.

6. Clinical communication. If RNA testing is pending while DNA results are released, the report should make the incomplete fusion assessment explicit.

This is where analytical validation becomes inseparable from clinical utility. A high sensitivity estimate has limited value if the laboratory cannot distinguish a valid expressed fusion from an artifact, cannot report negative results within the relevant timeframe, or cannot maintain performance in the FFPE material routinely received.

The laboratory must also avoid a common interpretive error: treating RNA positivity as self-authenticating evidence of therapeutic actionability. An expressed fusion may be biologically real but not necessarily oncogenic, in-frame, recurrent, or supported by the relevant clinical evidence. The report requires a hierarchy that separates detected transcript structure, analytical confidence, biological interpretation, and treatment relevance.

Resolving complex rearrangements in KRAS wild-type cancers

The value of RNA fusion analysis becomes particularly visible in tumors where a common driver is absent. Pancreatic cancer provides a useful example. In an analysis of 803 pancreatic cancer samples using RNA fusion-calling methods including Arriba, druggable driver fusions involving ALK, BRAF, FGFR2, NRG1, NTRK1, NTRK3, RET, and ROS1 were identified primarily in KRAS wild-type tumors.

The finding supports a rational stratification strategy: when a dominant driver is not present, the search for alternative actionable events becomes more important, not less. A KRAS wild-type result does not itself establish that a fusion exists, and it should not be used as a substitute for fusion testing. It does, however, identify a molecular context in which transcript-level interrogation may provide disproportionate value.

This pattern also illustrates why a panel’s gene list is only one part of the test. The laboratory must consider:

• Whether the assay captures known and novel partners for clinically relevant genes.
• Whether the design can detect exon-skipping or atypical transcript structures.
• Whether the output distinguishes an in-frame fusion from an out-of-frame or nonfunctional event.
• Whether the detection algorithm is calibrated for low-expression transcripts.
• Whether the reporting system links the molecular event to an appropriate evidence tier without overstating the treatment implication.

For NTRK fusion RNA validation, for example, the relevant question is not simply whether the assay detects a sequence involving NTRK1, NTRK2, or NTRK3. The laboratory must establish that its design can identify the partner diversity and transcript configurations expected in clinical material, while controlling for non-specific chimeric reads and biologically irrelevant rearrangements. Similar considerations apply to RET, ROS1, ALK, and other fusion-driven pathways.

RNA also helps resolve cases in which DNA sequencing reports a structural abnormality but cannot determine its functional consequence. A transcript-level finding can show whether the rearranged segments are expressed and joined at a plausible exon boundary. That evidence does not eliminate the need for interpretation, but it can convert an ambiguous genomic signal into a more clinically meaningful molecular result.

Reporting, regulation, and the next diagnostic paradigm

The integration of reflex RNA sequencing changes the definition of a negative molecular report. A report based solely on DNA sequencing should state the boundaries of its fusion detection capability, particularly when large intronic regions are incompletely represented. Otherwise, the clinical team may read a technically limited negative result as a comprehensive exclusion of actionable rearrangements.

This is a reporting issue as much as an assay issue. A clinically useful report should distinguish among:

• No actionable alteration detected within the interrogated DNA regions.
• No fusion detected by the validated DNA component, with RNA reflex testing pending or not performed.
• RNA quality insufficient for reliable fusion analysis.
• A fusion detected with adequate analytical support but requiring interpretation or confirmation.
• A fusion detected and classified according to the laboratory’s evidence framework.

Such distinctions preserve the boundary between an assay result and a biological conclusion. They also make the diagnostic pathway auditable. If a patient later receives a fusion-directed therapy based on an external result, the laboratory can determine whether the event was outside its assay design, missed because reflex testing was not triggered, or present but filtered during interpretation.

Regulatory and quality systems will increasingly need to address the combined behavior of DNA and RNA assays rather than evaluating them as isolated technologies. Validation should cover not only analytical sensitivity and specificity but also specimen adequacy, fusion partner diversity, transcript expression range, complex rearrangements, and the handling of discordant DNA-RNA results. The laboratory’s claims must remain aligned with the evidence generated during validation.

No universal threshold currently determines when reflex RNA sequencing becomes mandatory across all pathology laboratories. The optimal policy depends on tumor prevalence, available tissue, expected fusion burden, local treatment pathways, reimbursement conditions, and laboratory capacity. The absence of a universal threshold does not diminish the clinical argument; it means that implementation must be calibrated to the population being served.

The more durable conclusion is methodological. DNA sequencing and RNA sequencing do not compete for the same information. DNA provides a broad view of genomic variation, including alterations that may not be expressed. RNA provides direct evidence of expressed transcript architecture and is particularly effective when the clinical event is a fusion with a difficult or variable intronic breakpoint. A precision oncology workflow that treats one as a complete substitute for the other will systematically preserve blind spots.

Reflex RNA sequencing for gene fusions is therefore best understood as an extension of molecular resolution. It adds a second layer of evidence where DNA profiling is structurally constrained, improves actionable fusion detection in selected tumor types, and creates a more complete basis for targeted therapy selection. The operational cost is measurable—one evaluated NSCLC workflow added two working days—but so is the consequence of an incomplete biomarker assessment.

The forward trajectory is not toward RNA-only diagnostics. It is toward coordinated DNA-RNA profiling in which assay selection follows the biology of the suspected alteration, tumor-specific risk determines reflex rules, and negative results are interpreted according to actual analytical coverage. In that model, the molecular laboratory becomes less dependent on a single sequencing modality and more capable of matching the test to the biomarker. That is the more rigorous definition of precision medicine: not maximal sequencing, but appropriate resolution of the alteration that governs treatment.