Table of contents

• Gene fusions: what are they and their relevance in myeloid malignancies
• Conventional methods for fusion detection
• Targeted capture panels for fusion gene detection
• How the SOPHiA DDM™ Platform maximizes fusion detection in DNA-based workflows
• SOPHiA GENETICS’ CARDAMOM fusion calling workflow

Gene fusions: what are they and their relevance in myeloid malignancies

Gene fusions occur when two independent genes become abnormally joined, creating a hybrid gene. These events can result from missplicing at the RNA level or structural rearrangements at the DNA level due to damage and faulty repair processes. Such rearrangements often include chromosomal translocations, interstitial deletions, and inversions¹

The resulting fusion involves a driver gene and one or more partner genes, joined at specific breakpoints that give rise to different fusion isoforms. These hybrid genes often encode abnormal fusion proteins that can hijack key cellular processes and promote oncogenesis.

A landmark example is the Philadelphia chromosome in chronic myeloid leukemia (CML), a translocation between chromosomes 9 and 22 that produces the BCR::ABL1 fusion gene. This discovery transformed CML treatment and paved the way for targeted therapies such as imatinib, a tyrosine kinase inhibitor (TKI) that dramatically improved patient survival[1].

Since then, numerous other fusion genes involving tyrosine kinases have been identified in myeloproliferative malignancies, including:

• FIP1L1::PDGFRA
• PDGFRB fusions
• FGFR1 rearrangements

These findings define a subgroup of disorders now classified as myeloid neoplasms with eosinophilia and abnormalities in PDGFRA, PDGFRB, or FGFR1. But while fusions involving PDGFRA and PDGFRB respond well to imatinib, FGFR1-driven diseases do not, highlighting the critical need for accurate and timely fusion detection².

Beyond tyrosine kinase fusions, several other fusion genes play critical roles in blood cancers and carry important diagnostic and prognostic implications³:

• PML::RARA: Found in the vast majority of acute promyelocytic leukemia (APL) cases, this fusion was a breakthrough discovery. It led to the development of targeted therapies such as all-trans retinoic acid (ATRA) and arsenic trioxide (ATO), which transformed APL from one of the most fatal subtypes of acute leukemia into one with long-term survival rates exceeding 95%.
• MLL (KMT2A) fusions: These occur across acute myeloid leukemia (AML), B-cell precursor acute lymphoblastic leukemia (BCP-ALL), and, less commonly, T-cell ALL. MLL is often referred to as a "promiscuous" oncogene due to its ability to fuse with a large number of different partner genes. AML cases involving MLL fusions are typically associated with poor prognosis.
• CFB fusions: Core-binding factor (CBF) AML involves recurrent fusions such as RUNX1::RUNX1T1 and CBFB::MYH11, which disrupt normal hematopoiesis by interfering with the CBF transcription factor complex. Although these fusions are not direct drug targets, they are clinically important markers. CBF-fusion AML is typically associated with a favorable prognosis, with high remission rates and long-term survival. Patients typically respond well to standard induction followed by high-dose cytarabine consolidation, often avoiding the need for stem cell transplantation in first remission.
• NUP98 fusions: Initially identified in AML, NUP98 can fuse with over 30 different partner genes, many of which encode homeobox transcription factors. NUP98 rearrangements are implicated in a range of hematologic malignancies, including AML, T-cell ALL, early T-cell precursor ALL, and myelodysplastic syndromes (MDS), and are often linked to adverse outcomes.

Conventional methods for fusion detection

Traditional cytogenetic techniques, such as chromosomal banding, alongside molecular methods like fluorescence in situ hybridization (FISH), gene fusion microarrays, and PCR/RT-PCR, were among the earliest clinical assays developed to detect fusion genes.

Today, these approaches remain highly sensitive and are routinely used for orthogonal confirmation of fusion findings, helping ensure diagnostic accuracy. However, they remain hard to scale due to their inherent limitations:

• Target only one fusion at a time
• Unable to detect novel fusion genes or complex structural rearrangements
• Involve time-consuming, costly, and iterative workflows

While still indispensable in many diagnostic workflows, these methods are increasingly being complemented by next-generation sequencing (NGS) technologies.

NGS-based approaches have dramatically expanded the scope of fusion detection, uncovering previously undetectable events and enabling a more comprehensive view of the genomic landscape. When integrated with traditional methods, these advanced tools help streamline diagnostics and improve clinical decision-making in the context of myeloid malignancies⁵.

Targeted capture panels for fusion gene detection

Targeted capture panels based on DNA or RNA offer a focused and efficient approach to detecting gene fusions. These panels rely on custom-designed probes that target regions commonly involved in fusion breakpoints^1,2,5.

In DNA-based NGS, probes aim to capture intronic regions where fusion breakpoints often occur. In this case, probe design and bioinformatic analysis pipelines play a key role in the sensitivity of the detection. Since fusion breakpoints frequently lie in large intronic regions that are GC-rich, repetitive, and structurally complex, this can impact capture efficiency and analysis complexity.

Despite these challenges, DNA-based capture panels remain valuable for:

• Detecting rare fusion partners, intergenic breakpoints, and non-canonical events
• Offering low input requirements
• Allowing fusion detection alongside other variant types (e.g., SNVs, indels, CNVs) in a single workflow

RNA-based NGS, by contrast, offers several advantages. Rather than capturing introns, RNA methods detect exon–exon junctions, which directly indicate the presence of expressed fusion transcripts. This simplifies probe or primer design and improves detection in regions that are difficult to sequence from DNA. RNA-based approaches are generally more sensitive than DNA-based ones and are especially useful for identifying expressed, clinically relevant fusions, including those with unknown or variable breakpoints.

However, RNA-based fusion detection is not without challenges:

• RNA quality is critical. Degradation, particularly in FFPE samples, can reduce sensitivity and cause uneven coverage
• RNA is inherently less stable than DNA, making sample handling and extraction conditions more sensitive
• Expression-based detection may miss fusions that are present at the DNA level but not actively transcribed

Despite these challenges, RNA sequencing remains a powerful tool, often used in tandem with DNA-based methods to increase confidence in fusion calls. In practice, DNA- and RNA-based approaches are complementary. DNA sequencing enables broader biomarker profiling from stable material, while RNA provides functional confirmation and increased detection sensitivity. Together, they support a more comprehensive understanding of gene fusions in myeloid malignancies.

How the SOPHiA DDM™ Platform maximizes fusion detection in DNA-based workflows

DNA-based fusion detection within the SOPHiA DDM™ Platform is driven by a two-pronged strategy that combines optimized probe design with advanced algorithmic analysis:

• Targeted probe design: In our comprehensive SOPHiA DDM™ Community Myeloid Solution, probes are carefully engineered to capture both gene partners in clinically actionable fusions, such as BCR::ABL1, PML::RARA, and FIP1L1::PDGFRA. This ensures robust detection of well-characterized fusions, while still enabling the discovery of rare or novel events through partner-agnostic fusion calling when only one fusion partner is targeted by the application.
• Advanced bioinformatics: SOPHiA GENETICS’ algorithm, CARDAMOM, applies rigorous probabilistic models and signal quality filters to distinguish true fusions from background noise, increasing confidence in each reported event.

CARDAMOM pinpoints potential fusion events (also referred to as adjacencies) by analyzing two key signals in sequencing data:

• Split read: a single read that maps partially to two different genomic regions. This suggests it spans a fusion breakpoint — a strong direct signal.
• Discordant reads: paired-end reads that map further apart than expected, on different chromosomes, or in the wrong orientation. These indicate a potential structural rearrangement.

Once key read signals are detected, CARDAMOM clusters them into candidate breakpoints and applies a rigorous multi-step process to ensure only high-confidence fusions are reported. This fusion-calling workflow is designed to maximize sensitivity while maintaining analytical precision.

SOPHiA GENETICS’ CARDAMOM fusion calling workflow

1. Detects split and discordant read pairs from paired-end sequencing data that suggest structural variation.
2. Uses a two-dimensional uncertainty model to account for positional variation on both sides of the breakpoint.
3. Evaluates ≥8 features through a probabilistic multivariate model — including fragment support, mismatch rate, sequence complexity, and mapping quality — to estimate the likelihood of a true fusion event.
4. Applies filters such as ≥3 CUMIN® groups, probability ≥0.5, and on-target breakpoints, then reports validated fusions in VCF format, including variant fractions.

Discover how our comprehensive application, SOPHiA DDM™ Community Myeloid Solution, enables confident fusion detection in clinical samples.

Webinar: Evaluating Next-Generation Sequencing Solutions for Real-World Clinical Needs in Myeloid Malignancy → Watch the webinar

Visit our SOPHiA DDM™ Community Myeloid Solution page to learn more about this application.

References

1. Su X, Zheng Q, Xiu X, et al. Med-X. 2024;2:14.
2. Schröder J, Kumar A, Wong SQ. Methods Mol Biol. 2019;1908:125-138.
3. Gianfelici V, Lahortiga I, Cools J. Expert Rev Hematol. 2012;5(4):381-393.
4. Matsukawa T, Aplan PD, Stem Cells. 2020;38(11):1366-1374.
5. Heyer EE, Deveson IW, Wooi D, et al. Nat Commun. 2019;10:1388.