RNA fusion and splice detection

In one sentence

RNA fusion and splice detection tests which sequence joins are supported in captured RNA, without automatically establishing the underlying DNA change or protein consequence.

The intuition

Knowing how often a message appears is different from checking how its sentences were joined. A gene count measures abundance; a junction observation examines structure. The analogy stops at sequence: the assay must distinguish true molecular joins from ambiguous mapping and preparation artifacts.

How it works

Ribonucleic acid (RNA) splice junctions join retained transcript regions. Many are normal products of alternative splicing. A fusion transcript contains sequence assigned to different genes. A deoxyribonucleic acid (DNA) rearrangement can produce such a message, but an RNA junction does not itself establish the DNA mechanism. Transcription across neighboring genes can also produce joined messages; these have been observed in tumor and reference tissues. Nacu 2011.

A split read maps across a junction. A spanning pair has its two reads on opposite sides of a candidate fusion. Analysis evaluates support, alignment quality, similar sequences and the reference version. Repeated amplified copies are not independent original molecules. Fusion-caller benchmarks show that read length, expression and filtering affect detection; their optimal benchmark thresholds are not universal clinical cutoffs. Haas 2019.

Splice analysis can compare the fraction of eligible observations supporting alternative joins. A percent spliced in (PSI) estimate describes inclusion under a defined event and denominator. Local junction evidence may identify a changed choice without reconstructing a complete transcript. Unfinished nuclear RNA containing introns is not automatically pathological intron retention. Vaquero-Garcia 2016.

Evidence for a message's join leaves separate questions about its origin and consequence.

The assay must cover the join. Whole-transcriptome sequencing, short-end counting and targeted panels recover different information. Anchored multiplex polymerase chain reaction (PCR) can search from a targeted gene end toward a partner not known in advance. Its reach still depends on the selected targets and validated preparation. The original method included testing in formalin-fixed, paraffin-embedded (FFPE) specimens; that does not validate every later panel or junction. Zheng 2014.

Some fixed-RNA expression methods sequence ligated synthetic probes rather than original transcript-derived sequences. Their gene counts cannot be repurposed as unrestricted fusion or splice-sequence evidence. Official probe-chemistry documentation, version 7.2.

Why it matters in cancer

RNA can test whether a candidate DNA change has an expressed product. Mixed tissue still needs cell attribution. A supported junction may alter a protein's sequence or expression, but function, target dependence and clinical benefit require further evidence. Failure to detect a join is informative only within the assay's coverage and sensitivity.

Assay card

Field	What to retain
Measures and how	RNA joins: prepare a suitable library, sequence, align/assemble, count support and evaluate candidate-specific artifacts
Input and tissue cost	Protocol-compatible fresh, frozen or fixed material; extraction/library preparation consume allocated RNA or tissue; required input and RNA integrity are assay-specific
Output and units	Gene partners, reference/transcript version, junction coordinates and supporting reads or deduplicated molecules; PSI or another splice fraction requires its defined denominator
Thresholds	Validated support, coverage, mapping and detection criteria for the exact event class and specimen; no universal minimum read count
Failure modes	Poor RNA, uncaptured regions, low expression, similar sequences, duplicate support and library/assembly artifacts
Limits	No automatic DNA rearrangement, malignant-cell origin, intact protein, functional effect or drug benefit
Validation tier	Research calls differ from clinical fusion assays; verify the panel, event classes, specimen validation and confirmation policy; a caller name is not clinical validation

Common confusions

Gene expression versus junction evidence: total counts can hide altered products.
RNA fusion versus DNA fusion: a junction observation answers an RNA question.
Independent confirmation versus rerunning software: another analysis of the same reads does not create new molecular evidence.

Try it

A fictional RNA library has a candidate A–B junction supported by three identical amplified reads. The report calls it a confirmed DNA rearrangement and an effective drug target. What is missing?

Answer: Assess independent molecular support, coverage and artifacts first. DNA confirmation addresses the genomic mechanism; protein/function and disease-specific clinical evidence address the proposed treatment. Three copied reads cannot fill those gaps.

Explain it back

What can junction reads add beyond a gene count?

One possible answer: Evidence about a captured message's structure, with uncertainty about its source and downstream effect.

Takeaway

Authenticate the RNA join, then test the separate DNA, protein and clinical claims.

Sources and scope

Source check: October 10, 2026. The practice is fictional. Expert and learner review remain pending. Published method benchmarks do not establish a universal clinical threshold or a treatment recommendation.