The painstaking process of analyzing complex medical data is getting a significant boost from an unlikely source: generative artificial intelligence. Researchers at UC San Francisco and Wayne State University have demonstrated that AI can not only process enormous datasets far faster than traditional computer science teams, but in some instances, deliver even stronger results. This breakthrough in generative AI’s application to medical research could dramatically accelerate the pace of discovery, particularly in areas like preterm birth prediction.
The study, published February 17 in Cell Reports Medicine, involved challenging both human experts and AI systems to predict preterm birth using data from over 1,000 pregnant women. What emerged was a striking difference in efficiency. While human teams spent months meticulously analyzing the data, AI-assisted researchers – even a pairing of a UCSF master’s student, Reuben Sarwal, and a high school student, Victor Tarca – were able to generate functioning computer code in minutes, a task that would typically take experienced programmers hours or even days. This speed advantage is largely attributed to the AI’s ability to write analytical code based on concise, specialized prompts.
AI’s Unexpected Performance in Predicting Preterm Birth
The research wasn’t simply about speed; the AI’s performance was competitive. Of the eight AI chatbots tested, four produced usable code that matched the performance of the human teams, and in some cases, even surpassed it. “These AI tools could relieve one of the biggest bottlenecks in data science: building our analysis pipelines,” explained Marina Sirota, PhD, a professor of Pediatrics and interim director of the Bakar Computational Health Sciences Institute (BCHSI) at UCSF, and the principal investigator of the March of Dimes Prematurity Research Center at UCSF. “The speed-up couldn’t come sooner for patients who need aid now.”
The team’s approach involved leveraging a global crowdsourcing competition called DREAM (Dialogue on Reverse Engineering Assessment and Methods) which had already generated a wealth of data related to pregnancy and preterm birth. Researchers, led by Sirota and Adi L. Tarca, PhD, a professor at Wayne State University, instructed the AI systems to independently generate algorithms using these existing datasets, without direct human coding intervention. The AI chatbots were guided by carefully crafted prompts, similar to the way users interact with systems like ChatGPT, to analyze vaginal microbiome data and blood or placental samples to identify signs of preterm birth and estimate gestational age.
The Challenge of Preterm Birth and the Need for Faster Analysis
Preterm birth remains a significant public health concern. In the United States alone, roughly 1,000 babies are born prematurely each day, making it the leading cause of newborn death and a major contributor to long-term motor and cognitive challenges in children. Despite its prevalence, the underlying causes of preterm birth remain largely unknown. Sirota’s team compiled microbiome data from approximately 1,200 pregnant women across nine separate studies to investigate potential risk factors.
Analyzing such a vast and complex dataset presented a considerable hurdle. “This kind of operate is only possible with open data sharing, pooling the experiences of many women and the expertise of many researchers,” said Tomiko T. Oskotsky MD, co-director of the March of Dimes Preterm Birth Data Repository and associate professor at UCSF BCHSI. The DREAM competition had initially taken nearly two years to consolidate findings and publish results, highlighting the time-consuming nature of traditional data analysis methods.
From Competition to AI-Powered Acceleration
The researchers were curious if generative AI could significantly shorten this timeline. The partnership between Sirota’s group and Tarca’s team, who had previously led other DREAM challenges focused on estimating pregnancy stage, proved fruitful. The entire AI-driven effort – from initial instruction to paper submission – was completed in just six months, a dramatic reduction compared to the previous two-year process.
While the results are promising, scientists emphasize that AI is not a replacement for human expertise. The systems can produce misleading results and require careful oversight. However, by automating the initial stages of data analysis, generative AI can free up researchers to focus on interpreting results and formulating more meaningful scientific questions. “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 authors of the study include UCSF researchers Reuben Sarwal, Claire Dubin, Sanchita Bhattacharya, MS, and Atul Butte, MD, PhD, along with Victor Tarca (Huron High School, Ann Arbor, MI), Nikolas Kalavros and Gustavo Stolovitzky, PhD (New York University), Gaurav Bhatti (Wayne State University), and Roberto Romero, MD, D(Med)Sc (National Institute of Child Health and Human Development (NICHD)).
This work was funded by the March of Dimes Prematurity Research Center at UCSF and by ImmPort. The data used in the study was generated with support from the Pregnancy Research Branch of the NICHD.
Looking ahead, researchers plan to continue exploring the potential of generative AI in other areas of medical research. The next step involves applying these techniques to even larger and more complex datasets, with the goal of identifying new biomarkers and developing more effective interventions for a wider range of diseases.
Have thoughts on this story? Share your comments below.
