SYNTHEMA consortium partner, Humanitas, has made a significant contribution to a groundbreaking scientific paper recently published in the esteemed JCO Clinical Cancer Informatics, an American Society of Clinical Oncology journal. The paper, titled “Synthetic Data Generation by Artificial Intelligence to Accelerate Research and Precision Medicine in Hematology,” underscores the increasing role and potential of synthetic data in life sciences research.
Synthetic data, artificial constructs generated by an algorithm that learns from a real source data set, are being employed extensively to accelerate research. The paper details how generative artificial intelligence was applied to create synthetic data for different hematologic neoplasms.
Key to the study was the implementation of a conditional generative adversarial network architecture, generating synthetic data for conditions such as myelodysplastic syndromes (MDS) and Acute Myeloid Leukemia (AML). Furthermore, a robust validation framework was developed to assess the fidelity and privacy preservability of the synthetic data.
The results are compelling, revealing the creation of synthetic cohorts for MDS/AML with high fidelity and privacy performance. This new technology can effectively bridge gaps in information and enable data augmentation.
Interestingly, the study shows that starting with 944 patients with MDS from 2014, a 300% augmented synthetic cohort was created. This not only expedited the development of molecular classifications and scoring systems, but it also provided significant findings years earlier than they would have emerged using traditional research methods.
The synthetic data generated within the study successfully mirrored the clinical endpoints of a clinical trial with 187 MDS patients treated with luspatercept. The paper also announces the launch of a new website designed to enable clinicians to generate high-quality synthetic data from an existing biobank of real patients.
The paper concludes by underlining how synthetic data successfully mimic real clinical-genomic features and outcomes while preserving patient anonymity. It demonstrates the considerable potential of this technology to enhance the scientific use and value of real data, thereby accelerating precision medicine in hematology and clinical trial conduction.
This work represents a significant milestone in the integration of artificial intelligence within hematology and precision medicine. For more details and a thorough understanding of the research, we highly recommend reading the full paper at this link.