A new study in Cell Reports Medicine finds generative AI can process large-scale medical data far faster than conventional research teams, while delivering comparable predictive results.
Researchers conducted a direct comparison between human-only groups and teams deploying AI tools to analyse pregnancy data from more than 1,000 women.
The AI-assisted models frequently equalled and sometimes exceeded the performance of those created by investigators at UC San Francisco and Wayne State University.
One example involved a UCSF master’s student, Reuben Sarwal, and high school student Victor Tarca. Using generative AI, the pair produced working analytical code in minutes — a task that can take seasoned programmers days. They completed experiments, validated results, and submitted findings within months.
“These AI tools could relieve one of the biggest bottlenecks in data science: building our analysis pipelines,” said Marina Sirota, PhD, co-senior author of the study. “The speed-up couldn’t come sooner for patients who need help now.”
Preterm birth remains the leading cause of newborn death and contributes to long-term motor and cognitive challenges. Roughly 1,000 babies are born prematurely each day in the United States.
The researchers compiled microbiome data from about 1,200 women across nine studies. The datasets originated from three global DREAM challenges, where more than 100 teams developed predictive models. While competition work wrapped in three months, publishing the results took nearly two years, compared with the AI replication effort, which lasted six months.
According to the research, only four of eight AI systems tested generated usable code. Human oversight remained central.
“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,” said Adi L. Tarca, PhD. “They can focus on answering the right biomedical questions.”