AmpliconSuite checks its ecDNA calls against FISH on 175 genes in 72 cell lines
AmpliconSuite, a cancer-genome toolkit described in Nature, was validated against FISH on 175 genes in 72 models. The paper also creates AmpliconRepository, with more than 15,000 public results for labs to compare their own calls against.
The Scientist · Science desk
Two bars. 24 of 52 non-native HSR genes were marked neutral. The same 24 neutral-marked genes out of all 175 FISH-probed genes.
- Non-native HSR genes, neutral, claim 16 24 of 52 genes 46% computed
Non-native HSR genes, neutral The rest of the 52
- FISH-probed genes, neutral, claim 15 24 of 175 genes 14% computed
FISH-probed genes, neutral The rest of the 175
| Share | Part | Whole | Percent | Claim |
|---|---|---|---|---|
| Non-native HSR genes, neutral | 24 genes | 52 genes | 46% (computed) | 16 |
| FISH-probed genes, neutral | 24 genes | 175 genes | 14% (computed) | 15 |
What happened
- The FISH set held 56 ecDNA-borne genes, 3 ecDNA-indeterminate, 52 non-native HSR, 39 native HSR and 25 unknown HSR, and 90 of the 175 came from five earlier studies.
- AmpliconSuite-pipeline takes FASTQ, BAM or CRAM input, chains AmpliconArchitect with AmpliconClassifier, and is distributed through Bioconda and Nextflow.
- AmpliconArchitect v1.5.r2 runs 2.8 times faster at the median than v1.0, and 6.4 times faster on samples that originally took more than two hours.
- The authors also analysed 6,506 patient pan-cancer samples and report preferences in which oncogenes ecDNA co-amplifies.
Why it matters
- What's now possible capability A lab can set its own amplicon calls beside a public pool of results. The pool is community-contributed, so comparability depends on the contributors.
- What it limits constraint The FISH ground truth comes from cell lines, so accuracy on patient tumors has to be inferred from them.
- Who has to decide decision Teams choose whether to switch on the optional similarity-score filter that screens out reference-derived false positives, and the CCLE case shows what it caught.
The validation design decides what an accuracy figure would mean here. FISH images metaphase chromosomes, and ecDNA that has integrated into a chromosome typically appears as a non-native homogeneously staining region (HSR), meaning the amplified gene sits on a different chromosome than its centromere probe. The authors say complete integration is common in cancer cell lines [14]. Sequencing classified 24 of the non-native HSR genes as ecDNA, and the authors marked those cases neutral in the performance calculations [8]. That is 13.7% of the 175 genes [15] and 46.2% of the 52 non-native HSR genes [16].
The call is biologically reasonable. In CCF-STTG1, WGS found a single CDK6 ecDNA cycle, yet 3 of 19 metaphase images also showed non-native CDK6 HSRs, and the authors note that ecDNA-derived HSRs may coexist latently [9]. Still, a neutral case counts for neither side. Anyone benchmarking another caller on this FISH set would need to apply the same convention to get a comparable number.
The check in the other direction is cleaner but small. Eight of nine BFBs with a probed centromere were native HSRs [10]. Nine is a thin denominator.
The thing this doesn't tell you is how often the classifier is right. The abstract says the validation "demonstrated accurate predictions" [2], but the validation text breaks off before any sensitivity or precision figure appears. The speedups reported for AmpliconArchitect are likewise measured against its own v1.0 [7].
One false positive shows what the similarity filter is for. In Cancer Cell Line Encyclopedia data it revealed an identical ~61-kb amplified circular structure in three unrelated samples. The structure was a rare germline multi-copy duplication with a 2.2% population frequency, not ecDNA [11].
We think the evidence supports AmpliconSuite as a first-pass ecDNA classifier for cell-line WGS, where the FISH comparison was done [13]. For patient tumors, the cell-line set is indirect evidence at best.
What to watch
- Sensitivity and precision for the 175-gene validation, with the 24 neutral cases reported separately.
- Any FISH or other orthogonal check on patient tumor samples, beyond the cell-line set.
- How AmpliconRepository labels the AmpliconArchitect version behind each hosted result.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence62
- Adoption
- Not scored
- Hype gap+15
- Incentives
- Not scored
- Confidence58
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
AmpliconSuite is a toolkit for focal amplification analysis from cancer whole-genome sequencing that identifies ecDNA, breakage-fusion-bridges and other amplification modes.
- [2]
The abstract states that validation across 175 metaphase FISH-probed genes in 72 models 'demonstrated accurate predictions'.
- [3]
Analysis of 6,506 patient pan-cancer samples revealed ecDNA-mediated oncogene co-amplification preferences.
- [4]
AmpliconRepository is a community-contributable platform that enables investigation of results across more than 15,000 publicly hosted samples.
- [5]
The FISH set comprised 56 ecDNA-borne, 3 ecDNA-indeterminate, 52 non-native HSR, 39 native HSR and 25 unknown HSR genes, including 90 genes from five previous studies.
- [6]
AmpliconSuite-pipeline predicts focal amplifications from FASTQ, BAM or CRAM input, combines input preparation, AmpliconArchitect and AmpliconClassifier, and is available through Bioconda and Nextflow.
- [7]
AmpliconArchitect v1.5.r2 showed a 2.8 times median speedup over v1.0, increasing to 6.4 times for samples originally taking more than 2 hours.
- [8]
There were 24 non-native HSR genes with an ecDNA WGS classification, and the authors marked FISH:non-native HSR/AC:ecDNA cases as neutral in performance calculations.
- [9]
In CCF-STTG1, WGS identified a single CDK6 ecDNA cycle explaining the amplification, yet 3 of 19 metaphase images also showed non-native CDK6 HSRs; the authors note ecDNA-derived HSRs may coexist latently.
- [10]
Eight of nine BFBs with a probed centromere were native HSRs.
- [11]
In Cancer Cell Line Encyclopedia data, similarity-score filtering revealed an identical ~61-kb amplified circular structure in three unrelated samples, identified as a rare germline multi-copy duplication (2.2% population frequency) and not ecDNA.
- [12]
An optional similarity score-based filter controls reference-derived false positives by computing similarity scores between amplicons.
- [13]
The 175 FISH-probed genes were across 72 cell lines having closely matching WGS.
- [14]
Complete integration of ecDNA into chromosomes is common in cancer cell lines and typically appears as a non-native HSR, an amplified gene on a different chromosome than its centromere probe.
- [15]
The 24 neutral-marked genes are about 13.7% of the 175 FISH-probed genes.
- [16]
The 24 neutral-marked genes are about 46.2% of the 52 non-native HSR genes.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- nature.comAmpliconSuite: an end-to-end workflow for analyzing focal amplifications in cancer genomes
1 article · October 15, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- Cancer genomicsFollow
- Bioinformatics pipelinesFollow
- Extrachromosomal DNA (ecDNA)Follow