Grant: Project Grant

Many cancers are rare, and within certain tumor types, specific biomarkers and mutations can be even rarer. Therefore, conducting adequately powered analyses and achieving meaningful results can be increasingly difficult. Frequently, data must be pooled from multiple sources or regions for research to occur. However, current data legislations are often prohibitive, thus making it challenging to share data across provincial and territory borders. Synthetic data, which are artificially generated to mimic the statistical properties and structures of real data without revealing any patient information, offers a promising solution to current legal barriers and ethical challenges associated with data sharing. The overall objective of this study is to use machine learning and statistical models to generate and validate high quality synthetic data for specific oncology scenarios, namely: (1) sarcomas, which is a rare cancer; (2) ALK and ROS mutated lung cancers, which are rare genetic variants; and (3) rural patients, who are frequently marginalized. All three groups face significant barriers to treatment advances due to their persistent under-representation in clinical trials as well as their heightened privacy concerns due to their small sample sizes. Therefore, synthetic data may help accelerate research in these underrepresented populations by creating datasets that can be shared more easily across regions, combined together to improve sample sizes, and rebalanced to enhance their diversity and representativeness Rigorously generated and validated synthetic datasets hold immense potential to revolutionize research and expedite progress by addressing the critical issues of data sharing and privacy, scarcity, and diversity. The lessons gained from this proposed study can be leveraged to replicate our work in other jurisdictions, facilitate broader implementation for the use of synthetic data, and expand the reach and impact of our findings beyond Canada.