Scalable integration of multiomic single-cell data using generative adversarial networks
2024

Integrating Multi-Omic Single-Cell Data with MOWGAN

publication Evidence: high

Author Information

Author(s): Giansanti Valentina, Giannese Francesca, Botrugno Oronza, Gandolfi Giorgia, Balestrieri Chiara, Antoniotti Marco, Tonon Giovanni, Cittaro Davide

Primary Institution: IRCCS San Raffaele Scientific Institute

Hypothesis

Can a generative adversarial network effectively integrate multi-omic data from single cells without prior assumptions?

Conclusion

MOWGAN successfully integrates multi-omic data from single cells, revealing hidden biological patterns.

Supporting Evidence

  • MOWGAN can integrate more than two molecular layers without changing its architecture.
  • It effectively generates synthetic datasets that maintain the biological relevance of the original data.
  • MOWGAN outperforms existing tools in integrating unpaired multi-omic data.
  • Semisupervised training improves the quality of the generated data.
  • MOWGAN reveals hidden patterns in colorectal cancer organoids.

Takeaway

MOWGAN is like a smart helper that combines different types of information from tiny cells to help scientists understand them better.

Methodology

MOWGAN uses Wasserstein Generative Adversarial Networks to integrate multi-omic data from single cells.

Potential Biases

Potential biases may arise from the assumptions made during the integration process.

Limitations

MOWGAN's performance may vary based on the complexity of the datasets and the specific modalities used.

Digital Object Identifier (DOI)

10.1093/bioinformatics/btae300

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication