TL;DR: Researchers at UC San Francisco and Wayne State University found that generative AI tools could build medical prediction models dramatically faster than human teams — and in some cases produced more accurate results. The study focused on predicting preterm birth from data covering more than 1,000 pregnant women.

From Months to Minutes

The study, published in Cell Reports Medicine, compared human research teams working alone against those using AI assistance. All were given the same challenge: build machine learning models to predict preterm birth using microbiome data from roughly 1,200 pregnant women across nine studies.

Of eight AI chatbots tested, four produced usable code that generated prediction models matching or exceeding those built by experienced data scientists. A junior pair — a master’s student and a high school student — successfully built functional models with AI help, generating analytical code in minutes that would typically take experienced programmers hours or days.

The entire AI-assisted project, from concept to journal submission, took six months. The original human-only competition (DREAM challenge) took three months to complete the analysis and nearly two years to compile and publish findings.

A Bottleneck Broken

“These AI tools could relieve one of the biggest bottlenecks in data science: building our analysis pipelines,” said Marina Sirota, professor of Pediatrics and interim director of the Bakar Computational Health Sciences Institute at UCSF. “The speed-up couldn’t come sooner for patients who need help now.”

Preterm birth remains the leading cause of newborn death, with about 1,000 babies born prematurely every day in the US alone. Scientists still do not fully understand its triggers, and the complexity of the data involved has slowed research progress.

Looking Forward

The researchers stress that AI does not replace scientific expertise — human oversight remains essential, and AI systems can still produce misleading results. But the findings suggest AI could free researchers from debugging code and building pipelines, letting them focus on asking better scientific questions. For UK health research organisations, this points toward a practical use case where AI is already delivering measurable results rather than theoretical promise.