Show Notes
Díaz-Navarro A et al., Cell Genomics - OncoGAN is a multimodel generative AI pipeline that simulates realistic, privacy-preserving cancer genomes (VCFs, CNAs, SVs) across eight tumor types to support benchmarking and training of genome-analysis tools. Key terms: OncoGAN, synthetic genomes, generative AI, mutational signatures, privacy-preserving.
Study Highlights:
OncoGAN combines GANs, TVAEs, CTGAN/CTAB-GAN+ and sampling to model donor characteristics, mutational signatures, genomic positions, VAFs, CNAs and SVs and generates realistic synthetic cancer genomes. The pipeline reproduces tissue-specific mutational patterns and driver correlations while ensuring minimal overlap with training donors to preserve privacy. Augmenting DeepTumour training sets with OncoGAN-simulated donors improved classification accuracy, and the authors released 800 simulated genomes for community use.
Conclusion:
OncoGAN produces shareable, privacy-preserving synthetic cancer genomes that closely recapitulate real tumor features and can improve development and benchmarking of cancer-genome analysis tools.
Music:
Enjoy the music based on this article at the end of the episode.
Article title:
In silico generation of synthetic cancer genomes using generative AI
First author:
Díaz-Navarro A
Journal:
Cell Genomics
DOI:
10.1016/j.xgen.2025.100969
Reference:
Díaz-Navarro A, Zhang X, Jiao W, Wang B, Stein L. In silico generation of synthetic cancer genomes using generative AI. Cell Genomics. 2025;5:100969. https://doi.org/10.1016/j.xgen.2025.100969
License:
This episode is based on an open-access article published under the Creative Commons Attribution 4.0 International License (CC BY 4.0) – https://creativecommons.org/licenses/by/4.0/
Support:
Base by Base is independent and ad-free — no sponsors, no paywall. If an episode was worth your time, chip in and keep the papers audited and the original songs coming:
❤️ Support monthly: https://buy.stripe.com/cNifZhclVebvagk2JDgEg01
☕ One-time donation: https://donate.stripe.com/7sY4gz71B2sN3RWac5gEg00
More at basebybase.com
On PaperCast Base by Base you'll discover the latest in genomics, functional genomics, structural genomics, and proteomics.
Episode link: https://basebybase.com/episodes/in-silico-generation-of-synthetic-cancer-genomes-using-generative-ai
QC:
This episode was checked against the original article PDF and publication metadata for the episode release published on 2025-09-04.
QC Scope:
- article metadata and core scientific claims from the narration
- excludes analogies, intro/outro, and music
- transcript coverage: Audited core methodology and results sections describing the OncoGAN/OncoGAN-like Ankogan pipeline, privacy safeguards, genome discretization, mutational signatures, driver-detection performance, DeepTumour validation, data accessibility, and stated limitations.
- transcript topics: OncoGAN/Ankogan pipeline architecture (GANs, TVAEs, and random sampling); Privacy-preserving data generation (independent location and context generation; final assembly); Genome discretization and position mapping (3Gbp to ~30Mbp; binning and expansion); Mutational signatures and their recovery (SBS4, SBS9; SigProfilerExtractor validation); Driver mutation analysis with ActiveDriverWGS; DeepTumour tissue-of-origin prediction and impact of synthetic data
QC Summary:
- factual score: 10/10
- metadata score: 10/10
- supported core claims: 7
- claims flagged for review: 0
- metadata checks passed: 4
- metadata issues found: 0
Metadata Audited:
- article_doi
- article_title
- article_journal
- license
Factual Items Audited:
- OncoGAN uses a multimodel ensemble (GANs, TVAEs, random sampling) to generate eight-tumor-type synthetic cancer genomes.
- Privacy safeguards: mutation location and trinucleotide context are generated independently and combined at the end; 100 nt window check verifies context against real genome; only
- Genome discretization compresses 3,000,000,000 bp to 30,000,000 bp and bins into over 100,000 segments (with explicit mention of 105 bins of 0.3 Mbp in the canonical article).
- Mutational signatures SBS4 and SBS9 are reproduced in synthetic data; SigProfilerExtractor recapitulates the real donor distributions.
- ActiveDriverWGS detects ~89% of coding driver genes in real data and ~87% in synthetic OncoGAN data.
- DeepTumour achieves nearly 100% tissue-of-origin prediction accuracy on synthetic genomes for most tumor types; synthetic data improves training when mixed with real data.
QC result: Pass.
