AI Accelerates Medical Research: Predicting Preterm Birth Faster Than Ever Before

The pace of medical discovery may be accelerating thanks to advances in generative artificial intelligence. A latest study reveals that these systems can analyze complex medical datasets far more quickly than traditional research teams, potentially shortening the timeline from data collection to actionable insights. In one instance, a team of researchers – including a master’s student and a high schooler – developed functioning computer code in minutes using AI support, a task that would typically take experienced programmers hours or even days.

The research, a collaboration between scientists at UC San Francisco and Wayne State University, focused on predicting preterm birth, a leading cause of newborn death and long-term health challenges. Researchers assigned identical tasks to groups relying on human expertise alone, and to those using scientists working with AI tools. The findings, published in Cell Reports Medicine on February 17, suggest that generative AI could relieve a significant bottleneck in data science: the creation of analytical pipelines.

The Challenge of Preterm Birth Research

Roughly 1,000 babies are born prematurely in the United States each day, and the causes of preterm birth remain incompletely understood. To investigate potential risk factors, researchers compiled microbiome data from approximately 1,200 pregnant women across nine separate studies. Analyzing such a vast and complex dataset presented a considerable challenge, prompting the team to explore the potential of generative AI.

The study builds on previous function conducted through the DREAM (Dialogue on Reverse Engineering Assessment and Methods) project, a global crowdsourcing competition focused on machine learning in healthcare. Researchers had previously used DREAM challenges to develop machine learning models to detect patterns linked to preterm birth and to improve methods for estimating pregnancy stage. As reported by Today@Wayne, Adi Tarca, PhD, of Wayne State University, co-led one of the DREAM challenges focusing on improving methods for estimating pregnancy stage.

How Generative AI Accelerated Analysis

To assess the potential of generative AI, researchers instructed eight AI systems to independently generate algorithms using the same datasets from the DREAM challenges, without direct human coding. The AI chatbots received detailed, natural language instructions, similar to those used when interacting with systems like ChatGPT, guiding them to analyze the health data in a manner comparable to the original DREAM participants.

The AI systems were tasked with analyzing vaginal microbiome data to identify signs of preterm birth and examining blood or placental samples to estimate gestational age. Accurate pregnancy dating is crucial for appropriate prenatal care, as it determines the type of care a woman receives as her pregnancy progresses. Inaccurate estimates can complicate labor preparation.

While not all AI systems performed successfully – only four of the eight produced usable code – those that did often matched or even surpassed the performance of human teams. Crucially, these successful systems did not require large teams of specialists to guide them. The entire generative AI effort, from initial instruction to paper submission, took just six months, a significant reduction from the nearly two years it took to consolidate findings from the original DREAM project.

Beyond Speed: Focusing on Biomedical Questions

Marina Sirota, PhD, professor of Pediatrics and interim director of the Bakar Computational Health Sciences Institute (BCHSI) at UCSF, emphasized the potential of these tools to address a major bottleneck in data science. As ScienceDaily reports, Sirota stated that generative AI could allow researchers to spend less time troubleshooting code and more time interpreting results and formulating meaningful scientific questions.

“Thanks to generative AI, researchers with a limited background in data science won’t always demand to form wide collaborations or spend hours debugging code,” Tarca added. “They can focus on answering the right biomedical questions.”

Limitations and the Need for Oversight

Researchers are quick to point out that AI still requires careful oversight. These systems can produce misleading results, and human expertise remains essential for validating findings and ensuring accuracy. The study highlights that while AI can accelerate the analytical process, it does not replace the need for critical thinking and scientific rigor.

The success of only half of the AI chatbots as well underscores the importance of selecting appropriate tools and providing clear, specific instructions. The quality of the output is directly dependent on the quality of the input. Further research is needed to determine which AI systems are best suited for different types of medical data analysis and to develop best practices for prompting and validation.

The Future of Data-Driven Healthcare

This study represents an early, but promising, step towards integrating generative AI into healthcare research. The ability to rapidly analyze massive datasets could have implications for a wide range of medical fields, from drug discovery to personalized medicine. The researchers emphasize the importance of open data sharing, as demonstrated by the success of the DREAM project and the current study. Pooling data and expertise from multiple sources is crucial for advancing our understanding of complex diseases.

The researchers also acknowledge the funding support from the March of Dimes Prematurity Research Center at UCSF and ImmPort, as well as the data generated with support from the Pregnancy Research Branch of the NICHD. As reported by UC San Francisco, authors include Reuben Sarwal, Claire Dubin, Sanchita Bhattacharya, Atul Butte, Victor Tarca, Nikolas Kalavros, Gustavo Stolovitzky, Gaurav Bhatti, and Roberto Romero.

Looking ahead, the researchers plan to continue exploring the potential of generative AI in other areas of medical research. The focus will be on developing robust validation methods and ensuring that these tools are used responsibly and ethically to improve patient care. The next steps involve refining the prompting strategies and exploring different AI architectures to optimize performance and reliability.