AI chatbots aid pregnancy big data analysis

Marina Sirota, leader of the study, in her lab. Image by Elena Zhukova

A US study showed that artificial intelligence (AI) chatbots, when given the right prompts, can be used to decipher large amounts of health data faster than research teams. 

Researchers at UC San Francisco, California, collected data from 1,200 pregnant women, namely vaginal microorganisms, analysis of blood and placenta samples and birth dates. The data was analysed by several groups of data scientists, in a competition setting, to determine the stage of pregnancy and predict time of birth, which is important to determine the type of care that women should receive. Most groups took three months to analyse the large amount of data, but two years passed before they were ready to publish.  

Different teams of researchers, including one composed by a master student and a highschooler, used natural language to guide AI chatbots, including GPT-4o, Gemini 2.0 and DeepSeek-R1, to predict birth dates based on the same data. Four out of eight chatbots could prepare code that predicted birth dates as well as data scientists, sometimes outperforming human prediction of birth date.  

Teams using AI could run the experiments, confirm the results, write and submit their findings in just six months. While AI does not replace human expertise, since scientists need to safeguard against misleading results and takeover when AI-code fails, the study published in Cell Reports Medicine demonstrates the great potential of AI to support researchers.  

"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 Victor Tarca, co-author of the study from Huron High School, US. "They can focus on answering the right biomedical questions."