Article of the Week: Sepetmber 22

Synthetic data generation by artificial intelligence to accelerate research and precision medicine in hematology

Author

Marina Pulini-Franks MD
September 22, 2023

Synthetic Data Generation by Artificial Intelligence to Accelerate Research and Precision Medicine in Hematology

Synthetic data are artificial data generated without including any real patient information by an algorithm trained to learn the characteristics of a real source data set and became widely used to accelerate research in life sciences. We aimed to (1) apply generative artificial intelligence to build synthetic data in different hematologic neoplasms; (2) develop a synthetic validation framework to assess data fidelity and privacy preservability; and (3) test the capability of synthetic data to accelerate clinical/translational research in hematology.

ascopubs.org/doi/full/10.1200/CCI.23.00021?role=tab

Main Point

The article demonstrates the potential of generative artificial intelligence in creating synthetic patient cohorts with high clinical fidelity and privacy preservability, accelerating clinical and translational research in the field of hematology. Synthetic data offer a valuable solution to overcome the challenges of data collection, lack/incomplete information, and privacy concerns, enabling faster and more scalable access to representative information.

5 Salient Points

  • The study introduces a new technology based on generative artificial intelligence to generate synthetic patient cohorts with high clinical fidelity and privacy performances for personalized medicine in hematology.

  • Synthetic data efficiently replicate real clinical-genomic features and clinical outcomes, allowing data resolution for lack/incomplete information and data augmentation from real patients.

  • The technology's implementation in hematology shows generalizability across different clinical settings, including rare diseases with heterogeneous clinical and molecular backgrounds.

  • Synthetic data demonstrate the potential to accelerate translational research, anticipating and recapitulating insights from real-world cohorts in the field of personalized medicine.

  • The use of synthetic data as a comparison group in clinical trials offers the potential to accelerate clinical research, reduce costs, and eliminate patient concerns about treatment assignment, ultimately streamlining the conduction of clinical trials in hematology.

📝Research Appendix

Overview of experimental settings to validate synthetic data.

Definition of a molecular classification on augmented synthetic MDS cohort starting from 944 patients.