What is a fusion gene?
also called gene fusion · chimeric gene · translocation product
A fusion gene is what you get when two chromosomes break and the loose ends are glued back together in the wrong order, so the front half of one gene is stitched to the back half of another. The cell then reads this hybrid as one instruction and builds a hybrid protein that can do things neither original protein could.
The same thing, for a scientist
Gene fusions arise from structural rearrangements (translocations, inversions, deletions, insertions) that join the 5' portion of one gene to the 3' portion of another in-frame, producing a chimeric transcript and protein. In hematologic malignancies they are frequently initiating, clonal, and lineage-defining lesions; the KMT2A locus alone has >100 documented partner genes. Two qualifications: across pan-cancer data most detected fusions are passengers (fusions drive an estimated 16.5% of cases and are the sole driver in just over 1%), and recurrent chimeric transcripts - mostly read-through products of neighbouring genes - are a normal feature of non-cancerous human tissue, so a chimeric RNA is not by itself evidence of a rearrangement.
The kitchen analogy1 deeper
Think of a gene as a recipe card. A fusion is like tearing two cards in half and taping the top of the soup recipe to the bottom of the cake recipe. The kitchen (the cell) follows the taped card faithfully and produces something new. Whether that new thing is harmless or dangerous depends on which halves survived.
Why 'which half' matters
The 5' (front) half usually keeps the promoter (the on-switch) and any DNA-anchoring parts. The 3' (back) half contributes whatever machinery it carries. For KMT2A fusions the front half keeps the parts that dock onto DNA and onto the protein menin, while the back half (e.g., AFF1) brings a transcription-elongation engine. That combination is what locks blood stem cells in an immature, dividing state 3.
A short history
The first recurrent chromosome abnormality tied to a cancer was the 'Philadelphia chromosome' in chronic myeloid leukemia, shown in 1973 to be a reciprocal exchange between chromosomes 9 and 22 2. It later turned out to create the BCR::ABL1 fusion, the target of imatinib, the first drug designed against a fusion protein. That success is the template this project is trying to generalize 1.
How common are fusions in leukemia?
Fusions are found in a large share of childhood leukemias and are often the defining, earliest event. For the KMT2A gene alone, systematic sequencing of thousands of patients has catalogued well over 100 different partner genes, with a handful (AFF1, MLLT3, MLLT1, MLLT10, ELL, AFDN) accounting for the great majority of cases 3.
When a fusion is NOT a driver, and when a chimera is not a fusion
Two honest qualifications belong on this page, because the simple story above is the exception rather than the average case.
1. Most fusions found in cancers are passengers. Across 9,624 tumours, 25,664 fusions were called - and fusions were judged to drive only about 16.5% of cancer cases, and to be the sole driver in just over 1% [E1 counter, Gao 2018]. Leukaemia fusions like BCR::ABL1 and KMT2A::AFF1 are the clean, clonal, initiating cases that made the field; they are not representative of what a fusion caller returns.
2. Chimeric RNAs occur in healthy tissue. Nearly 300 RNA-seq libraries from 30 non-cancerous human tissues yielded many fusion transcripts, 291 of them recurrent; more than half of the recurrent ones join neighbouring genes transcribed in the same direction, using ordinary splice sites - so-called read-through or intergenic-splicing products. Some are functional in normal cells: silencing them reduced normal cell growth or motility [E1 counter, Babiceanu 2016].
The practical rule that follows: a fusion gene in the sense this page means requires a DNA-level rearrangement, clonality in the malignant population, and evidence that the chimeric product does something. A chimeric transcript alone satisfies none of those three.
| driver fraction pan cancer PCT | 16.5 |
|---|---|
| sole driver fraction pan cancer PCT | 1 |
| driver fraction source | Gao 2018, PMID 29617662 (9,624 tumours, 33 cancer types, 25,664 fusions called, 63% validation rate) |
| normal tissue chimeras | 291 recurrent chimeric fusion RNAs across 30 non-neoplastic human tissues; >50% join neighbouring same-strand genes (read-through / intergenic splicing). Babiceanu 2016, PMID 26837576 |
| kmt2a partner count in frame | 107 |
| kmt2a partner source caveat | Single referral-based diagnostic centre, 3401 patients 2003-2022 (Meyer 2023, PMID 37019990); no independent partner-stratified cohort published |
- Are there fusion transcripts in normal blood cells that the immune system or RNA surveillance tolerates, and what distinguishes those from oncogenic ones?
- Given that recurrent chimeric transcripts exist in normal tissue and that most cancer fusions are passengers, what is the minimum evidence this graph should require before giving a fusion its own entry? A stated threshold - DNA-level rearrangement plus clonality plus a functional readout - would be cheap to adopt and would make the atlas auditable.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.review
Balanced translocations and gene fusions are causal, often initiating, events in a large fraction of hematologic cancers.
The impact of translocations and gene fusions on cancer causation · Mitelman F, Johansson B, Mertens F · Nature Reviews Cancer 2007What this evidence cannot show
Review-level synthesis; individual fusions differ in whether they are sufficient alone.
Evidence that argues against this (2)
- Most detected fusions are not drivers. The TCGA Fusion Analysis Working Group called fusions across 9,624 tumours and 33 cancer types with multiple tools, identified 25,664 fusions at a 63% validation rate, and concluded that fusions drive the development of 16.5% of cancer cases and act as the SOLE driver in more than 1% of them. The passenger fraction is therefore very large. 'A fusion was found' is not the same claim as 'a fusion is driving this cancer', and the difference matters most in solid tumours; the leukaemia picture in this entry is the favourable end of the distribution, not the typical one. Gao Q et al. 2018
- Chimeric transcripts are a normal feature of normal cells. Babiceanu et al. curated nearly 300 RNA-seq libraries covering 30 non-neoplastic human tissues and cell types (plus 15 mouse tissues), found a large number of fusion transcripts, and identified 291 that recurred in more than one sample; RNA- and protein-level validation was performed on a subset. Over half of the recurrent fusions joined NEIGHBOURING genes transcribed in the same direction - read-through or 'intergenic splicing' products using canonical splice sites - and silencing several of the widely expressed ones reduced normal cell growth or motility, i.e. some of these chimeras do physiological work. So a chimeric RNA is not by itself evidence of a chromosomal lesion, and a fusion detected by RNA-seq needs a DNA-level rearrangement and a driver argument before it counts as what this entry describes. Babiceanu M et al. 2016
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The Philadelphia chromosome in CML is a reciprocal t(9;22) translocation.
A new consistent chromosomal abnormality in chronic myelogenous leukaemia identified by quinacrine fluorescence and Giemsa staining · Rowley JD · Nature 1973What this evidence cannot show
Historical; the molecular fusion (BCR::ABL1) was characterized a decade later.
Evidence that argues against this (1)
- True of the classic case and not of the entity as diagnosed. The Philadelphia chromosome was described in 1960 as a minute marker chromosome, years before banding revealed it as a translocation, and a minority of CML is Ph-NEGATIVE by cytogenetics while carrying BCR::ABL1 by molecular methods through cryptic or variant rearrangements. A definitional entry should say that the karyotypic appearance and the fusion gene are separable — which is the same lesson this graph records for KMT2A::SEPTIN11, cytogenetically indistinguishable from KMT2A::AFF1. Nature 1973
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The KMT2A locus has >100 documented fusion partners; a small set of partners dominates in infant and pediatric leukemia.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
Referral-based cohort (Frankfurt Diagnostic Center of Acute Leukemia); partner frequencies may be biased by which centers refer samples.
Evidence that argues against this (1)
- The >100-partner figure and the partner frequency ranking come from one referral-based diagnostic centre. The 2023 recombinome reports 3401 patients analysed between 2003 and 2022 at the Frankfurt Diagnostic Center of Acute Leukemia: 107 in-frame KMT2A fusions, the seven commonest partners plus PTD accounting for more than 90% of recombinations, 37 partners recurrent and 63 seen exactly once. No independent cohort has published a comparable partner-stratified distribution, so the long tail and the ranking both inherit this one centre's referral structure. The number is the best available; it is not an independently replicated one. Meyer C et al. 2023
builds on the fusion-gene explainer
- ← related toIs KMT2A::AFF1 leukemia epithelial or endothelial? No — it is haematopoietic. But the question is a good one
companion layman explainer
Counter-evidence (Babiceanu 2016, PMID 26837576): 291 recurrent chimeric fusion RNAs occur across 30 non-neoplastic human tissues, over half of them read-through products of neighbouring same-strand genes, so a chimeric transcript does not imply a chromosomal rearrangement
Counter-evidence (Gao 2018, PMID 29617662): across 9,624 tumours fusions drive only ~16.5% of cases and are the sole driver in just over 1%, so fusion detection is a weak basis for basket eligibility without a driver argument
the physiological counterpart to the pathological fusion explainer
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
the plain-English entry point
- related to →KMT2A / MLL1 — the gene that breaks
worked example of a fusion parent gene
a provenance refinement of the basic fusion concept
a worked example of the reader-plus-effector fusion architecture, in a family with no relation to KMT2A
layman entry point
a fusion family defined by the COMPLEXES its halves belong to rather than by a shared gene - a different kind of family from every one this graph names after a gene
A data-provenance caution attached to the fusion-biology entry point: the cell line an experiment used is metadata, it can be recorded inconsistently in a public deposit, and the filename every pipeline reads may carry no answer at all.
a measurement-kind caveat on fusion target claims
- ← related toA promoter window is not a gene: what it means to say a protein 'sits on gene X's switch'
companion explainer: what a promoter-occupancy number about a fusion oncoprotein does and does not assert
a worked case of how fusion architecture should and should not be described
a fusion gene can donate either end, and which end decides the druggable interface
a fusion-oncoprotein entity