Gene fusions—chimeric genes formed when two previously independent genomic segments are abnormally joined—are potent drivers of oncogenesis and critical targets for precision medicine. Historically, identifying these rearrangements relied heavily on DNA sequencing. However, identifying a structural break in DNA does not guarantee that it has any biological consequence. To determine whether a genomic rearrangement actually drives disease, researchers have shifted toward transcriptome-based fusion detection. By sequencing RNA rather than DNA, this approach captures the functional, expressed products of genomic instability.
The Intron Problem in DNA Sequencing
To understand why transcriptome-based detection works so well, we must first look at why DNA-based sequencing frequently struggles. Gene fusions typically occur due to large-scale chromosomal translocations, inversions, or deletions. The physical breakpoints of these rearrangements almost always land within introns—the vast, non-coding regions between exons.
Introns are notoriously difficult to sequence. They are massive and highly repetitive, making it technically challenging to design effective DNA capture probes. As noted by C. Xu et al. in Expert Consensus on the Diagnosis and Treatment of NRG1/2 Gene Fusion Solid Tumors, DNA-based sequencing assays often suffer from inadequate probe coverage across these expansive intronic spaces, leading to missed fusions.
Furthermore, even when DNA sequencing successfully uncovers an intronic breakpoint, it cannot determine the functional outcome of that break. Many structural rearrangements are benign "passenger" mutations that are never transcribed, or they result in non-functional intergenic fusions. As argued by Qing Yi et al. in Case Report: A case of first-line treatment for rare ROS1 fusion mutation lung adenocarcinoma with entrectinib, DNA sequencing can misidentify complex genomic rearrangements as non-functional, whereas RNA sequencing provides the transcriptomic ground truth by revealing whether an actionable, functional transcript is actually being produced.
Bypassing the Noise via Spliced Transcripts
Transcriptome sequencing (RNA-seq) bypasses the challenges of intronic sequencing by leveraging the cell’s own biology. Before an RNA molecule is translated into a protein, the cellular machinery splices out the introns, bringing the coding exons directly adjacent to one another.
DNA: [ Exon A1 ] --[ Massive Intron ]-- [ Breakpoint ] --[ Massive Intron ]-- [ Exon B2 ]
│
Splicing Machinery
▼
mRNA: [ Exon A1 ][ Exon B2 ] <-- Clean, continuous chimeric junction
By sequencing the mature mRNA, researchers can look directly at the continuous chimeric junction where the two genes meet. As demonstrated by Nike Beaubier et al. in Clinical validation of the tempus xT next-generation targeted oncology sequencing assay, this approach circumvents the technical hurdles of tiling across repetitive genomic regions, providing a highly sensitive and direct assessment of expressed fusions. This makes RNA-seq an ideal tool for mapping active oncogenic drivers across diverse tumor types, a strategy pioneered in early deep-sequencing surveys such as the one by David Jones et al. in Dissecting the genomic complexity underlying medulloblastoma.
The Computational Signals of a Fusion
To find these chimeric transcripts within millions of short RNA-seq reads, specialized bioinformatic alignment algorithms search for two distinct structural signals:
- Split Reads: A single sequencing read captures the exact boundary where the two genes fuse. The front half of the read maps perfectly to Gene A, while the back half maps to Gene B. The algorithm must "split" the alignment across different genomic loci, as detailed by Yufang Tang et al. in Transcriptomic insights into adenoid cystic carcinoma via RNA sequencing.
- Discordant Read Pairs: In paired-end sequencing, fragments are sequenced from both ends. If one end of a pair maps cleanly to Gene A and its partner end maps to Gene B, the pair is flagged as discordant. This indicates that the underlying transcript bridges the two genes, as explained by Meng Fan et al. in Case Report: A Pancreatic Ductal Adenocarcinoma Patient With Concurrent Targetable Somatic Novel KANK1-ALK, UPP2-NTRK3 Fusion, and Pathogenetic Germline BRCA Mutation.
To distinguish true biological fusions from sequencing artifacts or random transcriptional noise, integrated pipelines like the one evaluated by Shengyu Zhou et al. in Novel insights into molecular patterns of ROS1 fusions in a large Chinese <scp>NSCLC</scp> cohort: a multicenter study enforce a strict minimum threshold of unique split and discordant read pairs before confirming a fusion call.
Predicting Oncogenic Potential from Structure
Detecting a chimeric sequence is only half the battle. Standard transcriptome reconstruction tools can fail to identify complex fusions (Ting Yu et al. in TransBorrow: genome-guided transcriptome assembly by borrowing assemblies from different assemblers), and circular RNA detection tools often overlook fusions that cross gene boundaries (Quan Zou 0001 et al. in A comprehensive overview and evaluation of circular RNA detection tools). More importantly, a fusion must be evaluated to determine if it is an active oncogenic driver.
For a fusion protein to drive cancer, it must typically remain in-frame so that it can be translated into a fully functional, abnormal protein. If the genomic break shifts the reading frame, it will introduce premature stop codons, resulting in a degraded, non-functional transcript. As emphasized by Misako Nagasaka et al. in Pan-tumor survey of ROS1 fusions detected by next-generation RNA and whole transcriptome sequencing, highly sensitive RNA testing is required to confirm that a rearrangement yields an in-frame transcript capable of producing a disease-driving protein.
Advanced bioinformatic tools solve this by moving beyond simple sequence matching to perform deep structural analysis. The authors of FuGePrior: A novel gene fusion prioritization algorithm based on accurate fusion structure analysis in cancer RNA-seq samples (ELISA FICARRA et al.) demonstrated that evaluating structural features—such as the preservation of reading frames and upstream promoter regions—allows algorithms to filter out benign passenger mutations and technical noise. Furthermore, by computationally reconstructing the full chimeric protein sequence from the transcriptomic data, pipelines can verify whether critical functional domains (such as tyrosine kinase domains) are preserved intact (Francesco Abate et al. in Pegasus: a comprehensive annotation and prediction tool for detection of driver gene fusions in cancer).
Cross-Omic Validation
While RNA sequencing provides an exceptional window into functional expression, it is not entirely immune to errors. For example, Arthur Dondi et al. in Detection of isoforms and genomic alterations by high-throughput full-length single-cell RNA sequencing in ovarian cancer point out that sequencing methods that only capture the ends of transcripts can mistake a gene fusion for simple gene overexpression.
To eliminate these false positives, modern genomic workflows utilize cross-omic structural validation. By pairing transcriptomic sequencing with physical DNA mapping or long-read sequencing, researchers can confirm that a detected chimeric transcript is anchored by a true physical break in the genome.
- Single-Cell Multi-Omics: Integrating single-cell RNA sequencing with single-cell DNA sequencing allows for the precise validation of genomic breakpoints, ensuring that transcriptomic fusions are supported by physical DNA rearrangements rather than random transcriptional noise (Arthur Dondi et al.).
- Long-Read Integration: Combining long-read DNA sequencing with full-length transcriptomic sequencing allows researchers to map across highly repetitive regions of the genome, verifying that structural breaks directly cause the expression of novel proteins (Taobo Hu et al. in Detection of Structural Variations and Fusion Genes in Breast Cancer Samples Using Third-Generation Sequencing).
- Optical Genome Mapping: Integrating whole-transcriptome data with high-resolution physical maps of the genome confirms whether large-scale structural variations alter gene expression patterns (Yanan Cheng et al. in Optical Genome Mapping Reveals the Landscape of Structural Variations and Their Clinical Significance in HBOC-Related Breast Cancer).
This dual-layer validation strategy mirrors long-standing principles in genomics, such as using expressed sequence tags to distinguish active, transcribing transposable elements from degraded genomic fossils (Claudio Jose Struchiner et al. in Novel transposable elements from Anopheles gambiae). By combining the structural confirmation of DNA with the functional proof of RNA, cross-omic validation ensures that identified gene fusions are both structurally real and biologically active.
Key Foundational Work
Dissecting the genomic complexity underlying medulloblastoma (2012) — contributed 1 concept elements to this direction.
Novel transposable elements from Anopheles gambiae (2011) — contributed 1 concept elements to this direction.