Grateful to my colleagues and to my supervisors, Lincoln Stein & Bo Wang, for their guidance and support. Stay tuned —more to come!

Thrilled to see my postdoctoral work published in @cellpress.bsky.social

OncoGAN generates simulated genomes to train genomic analysis tools —without the confidentiality risks of real genomes.

News story: t.co/J9QJZInPOE
Paper: t.co/ygEjM5vuGZ

#Genomics #Cancer #AI #Bioinformatics

🚀 Only 2 weeks left to join the VI Visualization Contest with R by Grupo de R de Asturias!

🏆 Prizes:
🥇 1st: 300€
🥈 2nd: 100€

Show off your #RStats skills and impress us with your best visualizations!

🔗 More info: github.com/grupoRasturi...

#Visualization #Contest #DataViz

GitHub - LincolnSteinLab/oncoGAN: A pipeline that accurately simulates high quality publicly cancer genomes (VCFs, CNAs and SVs). A pipeline that accurately simulates high quality publicly cancer genomes (VCFs, CNAs and SVs). - LincolnSteinLab/oncoGAN

8/8 More info:

Alongside the OncoGAN models and pipeline, we’ve released 800 synthetic genomes spanning 8 tumor types!

A huge thank you to all the authors for their contributions to this work!!!

📄 Preprint: tinyurl.com/yepheye3
📂 Datasets: tinyurl.com/28bpd5hs
💻 Code & Docs: tinyurl.com/mr3ku653

7/8 Is OncoGAN useful? Absolutely!

- We tested ActiveDriverWGS on synthetic genomes to see if it could detect the same driver genes as in real patient data, proving its value in refining algorithms and defining detection limits.

6/8 Is OncoGAN useful? Absolutely!

- We used OncoGAN simulations to augment DeepTumour’s training dataset (a tool for identifying tumor type based on somatic mutation patterns), showing performance improvements.

5/8 What does OncoGAN simulate?

- Copy number alterations (CNA) and structural variants (SV): This updated version successfully simulates CNAs and SVs.

4/8 What does OncoGAN simulate?

- Tumor heterogeneity (A): Simulating donors with varying mutational burdens and characteristics.

- Tissue-specific mutational patterns (B): Accurately modeling the genomic distribution of mutations and mutational signatures unique to different tumor types.

3/8 Why is OncoGAN necessary?

- Benchmarking: Since the ground truth of real cancer genomes is often unknown, evaluations typically compare methods, introducing potential bias. By generating open-access synthetic genomes with a known ground truth, OncoGAN helps improve and benchmark these tools.

2/8 Why is OncoGAN necessary?

- Improving data sharing: We have demonstrated that OncoGAN does not leak any private patient data from its training set, a crucial factor given the sensitivity of genetic information as protected health data.

Our updated version of OncoGAN is out! 🚀

🧬 OncoGAN is an AI system capable of generating high-fidelity, open-access synthetic cancer genomes.

Do you want to know more about it? 1/8 🦋

Please consider spending a few moments of your day supporting our resource by providing your feedback to our team! The things we learn from the user survey is essential for our continued success!

🖥️🧬

We recently announced a dual new release of intOGen and boostDM

Computational analysis of 33,218 tumor genomes to identify cancer genes and driver mutations

➡️ Compendium of Cancer Driver Genes - www.intogen.org

➡️ In Silico Saturation Mutagenesis of Cancer Genes - www.intogen.org/boostdm