In a groundbreaking real-world test of artificial intelligence in healthcare, researchers from UC San Francisco and Wayne State University found that generative AI can analyze complex medical data significantly faster than traditional human research teams—sometimes even outperforming them. The study, published in Cell Reports Medicine on February 17, focused on predicting preterm birth using data from over 1,000 pregnant women, a critical health issue that remains poorly understood. Human teams had previously spent months analyzing the same data. In this new experiment, researchers compared results from human-only teams with those using AI tools. A junior team composed of a UCSF master’s student, Reuben Sarwal, and a high school student, Victor Tarca, successfully built predictive models with AI assistance. The AI generated functional code in minutes—tasks that typically take experienced programmers hours or days. The speed advantage came from AI’s ability to write analytical code based on concise, detailed natural language prompts. Out of eight AI chatbots tested, only four produced usable models, but those that succeeded required minimal human oversight. This allowed the junior team to complete their work, validate results, and submit their findings to a journal within just a few months—far faster than the nearly two years it took to publish earlier DREAM challenge results. Marina Sirota, PhD, a professor of Pediatrics at UCSF and interim director of the Bakar Computational Health Sciences Institute, co-led the study. She emphasized that AI could alleviate a major bottleneck in data science: building analysis pipelines. “The speed-up couldn’t come sooner for patients who need help now,” she said. Preterm birth is the leading cause of newborn death and a major contributor to long-term developmental challenges. In the U.S., about 1,000 babies are born prematurely each day. To better understand risk factors, Sirota’s team compiled vaginal microbiome data from approximately 1,200 pregnant women across nine studies. The research was part of the DREAM (Dialogue on Reverse Engineering Assessment and Methods) pregnancy challenges, a global competition involving over 100 teams. Most teams completed their work within three months, but consolidating and publishing findings took nearly two years. Sirota and Adi L. Tarca, PhD, professor at Wayne State University and co-senior author, decided to test whether generative AI could accelerate this process. They tasked eight AI systems with independently generating algorithms using the same datasets from the three DREAM challenges, without direct coding from humans. The AI systems were given precise natural language instructions, similar to how users interact with tools like ChatGPT. Their goals included identifying patterns in microbiome data linked to preterm birth and estimating gestational age using blood or placental samples—critical for determining appropriate prenatal care. After running the AI-generated code, researchers found that four systems produced models comparable to or better than those from human teams. The entire AI-driven research cycle, from start to paper submission, took only six months. While AI showed remarkable potential, researchers stress the need for human oversight. AI can generate misleading or inaccurate results, and expert judgment remains essential. Still, by rapidly processing vast datasets, AI enables researchers—especially those with limited data science training—to focus on scientific questions rather than technical hurdles. “Thanks to generative AI, researchers with a limited background in data science won’t always need to form wide collaborations or spend hours debugging code,” Tarca said. “They can focus on answering the right biomedical questions.” The study was funded by the March of Dimes Prematurity Research Center at UCSF and ImmPort, with data support from the Pregnancy Research Branch of the National Institute of Child Health and Human Development.