A team of researchers, led by MIT and Dana-Farber Cancer Institute, has used artificial intelligence (AI) to examine how thousands of medical studies share code. The study, recently published in Nature Medicine, aimed to investigate the current state of code sharing in clinical prediction studies. These studies develop or evaluate models to estimate patient outcomes, such as the risk of developing a disease. The researchers used large language models (LLMs) to analyze nearly 4,000 papers and identify statements about shared code. They then assessed hundreds of associated code repositories to evaluate the quality of the shared code and documentation. The review revealed two key gaps: code sharing remains uncommon, and making code public does not consistently mean providing the necessary documentation and supporting information for reuse. The findings are informing the development of TRIPOD-Code, a new reporting guideline for transparent code sharing. The guideline aims to provide a standardized framework for researchers to share their code and documentation, enabling others to scrutinize and build upon their work. The study highlights the importance of transparent code sharing in medical research. By making code and documentation publicly available, researchers can ensure that their findings are reproducible and reliable. This, in turn, can improve patient care and outcomes. The researchers' use of AI to analyze a large number of papers and code repositories demonstrates the potential of AI in facilitating transparent and reproducible research. The study's findings and the development of TRIPOD-Code are expected to have a significant impact on the medical research community. By promoting transparent code sharing, researchers can accelerate the discovery of new treatments and improve patient care. The study's results also underscore the need for researchers to prioritize transparency and reproducibility in their work.
Science
AI helps researchers investigate how thousands of medical studies share code
Prepared with AI assistance; not independently fact-checked. Original source: MIT News