In a groundbreaking study, researchers have demonstrated that generative AI can significantly enhance the speed of medical data analysis, particularly in predicting preterm birth. By assigning identical tasks to various groups, they compared the performance of teams that relied solely on human expertise with those that utilized AI tools alongside human scientists. The study focused on data from over 1,000 pregnant women.
Remarkably, even a junior research duo, consisting of UCSF master's student Reuben Sarwal and high school student Victor Tarca, successfully created predictive models with the assistance of AI. The generative AI system was capable of producing functional computer code in mere minutes--an endeavor that typically requires experienced programmers several hours, if not days, to complete.
The speed advantage of AI stems from its ability to generate analytical code based on concise and specific prompts. While not all AI systems excelled, four out of the eight chatbots produced usable code without needing extensive guidance from specialists. This efficiency enabled the junior researchers to finalize their experiments, validate their results, and submit their findings to a journal within a few months.
Marina Sirota, PhD, a Pediatrics professor and interim director of the Bakar Computational Health Sciences Institute at UCSF, highlighted the potential of AI tools to alleviate significant bottlenecks in data science. "The acceleration couldn't come at a better time for patients in need," she stated. This research was published in Cell Reports Medicine on February 17.
The Importance of Preterm Birth Research
Enhancing data analysis speed could lead to improved diagnostic tools for preterm birth, a leading cause of neonatal mortality and a significant factor in long-term developmental challenges for children. In the U.S., approximately 1,000 babies are born prematurely each day.
Despite ongoing research, the precise causes of preterm birth remain elusive. To explore potential risk factors, Sirota's team gathered microbiome data from about 1,200 pregnant women across nine studies. "This research is only feasible through open data sharing, combining the experiences of numerous women and the expertise of various researchers," noted Tomiko T. Oskotsky, MD, co-director of the March of Dimes Preterm Birth Data Repository.
Analyzing such complex datasets posed challenges, leading the researchers to engage in a global crowdsourcing initiative known as DREAM (Dialogue on Reverse Engineering Assessment and Methods). Sirota co-led one of the DREAM challenges, which focused on vaginal microbiome data, attracting over 100 teams worldwide to develop machine learning models that identified patterns associated with preterm birth.
AI's Role in Pregnancy and Microbiome Analysis
To assess whether generative AI could expedite the research timeline, Sirota's team collaborated with researchers led by Adi L. Tarca, PhD. Together, they instructed eight AI systems to autonomously generate algorithms from the same datasets used in the DREAM challenges, utilizing carefully crafted natural language prompts.
The AI systems aimed to analyze vaginal microbiome data for preterm birth indicators and assess blood or placental samples for gestational age estimation. Accurate pregnancy dating is crucial for determining appropriate care as pregnancies progress. The AI-generated models matched or even surpassed the performance of human teams, with the entire process from inception to publication taking only six months.
While AI shows promise in streamlining research, scientists emphasize the need for careful oversight, as these systems can yield misleading results. Nonetheless, generative AI may enable researchers to focus on interpreting findings rather than troubleshooting code, allowing for more meaningful scientific inquiries.
Thanks to generative AI, researchers can now concentrate on answering critical biomedical questions without the need for extensive collaborations or lengthy debugging sessions.
