There’s a new artificial intelligence (AI) assistant in town: ChatPTB.
Over the past year, while the world was busy using chatGPT to answer questions, make slide presentations and type out audio transcripts, researchers at the March of Dimes Prematurity Research Center (PRC) at the University of California, San Francisco (UCSF) were hard at work using it, too: to build an AI assistant focused on preterm birth.
Naturally, they named it ChatPTB, or ChatPreTermBirth.
ChatPTB, an AI agent that uses ChatGPT and other technologies to answer questions about research articles that have come out of March of Dimes PRCs, launched quietly in beta mode this past summer. It's housed in The March of Dimes Database for Preterm Birth Research, a one-of-a-kind multi-omic data bank administered by UCSF PRC scientists. Aimed at introducing the broader research community to March of Dimes data and science, ChatPTB is the first preterm birth research AI assistant in the field. It was jointly developed with colleagues at Peraton, a technology firm specializing in health informatics and security.
ChatPTB will respond to a wide range of questions that users can ask, like “what are the risk factors for preterm birth?”, “what are some therapeutic targets for preterm birth?”, “what is the most beneficial vaginal microbiome bacteria?”, and more. In addition to immediate responses, the AI agent also lists the scientific papers the answers were pulled from.
“We hope the benefit of this tool over ChatGPT will be greater accuracy of information,” said Dr. Marina Sirota, the UCSF PRC’s principal investigator. “And we hope that reliability will foster the type of scientific collaboration that will propel our research, and our common goal of ending preventable preterm birth, forward.”
One of the key reasons ChatPTB may deliver more accurate responses to user prompts than ChatGPT is because it's trained only on scientific papers funded by March of Dimes—not the entirety of content available on the internet. Plus, the agent has safeguards against making mistakes that can lead to AI “hallucinations,” or nonsensical responses.
“For scientists outside our field interested in collaboration, the biggest gift we could give them is clear, accurate information, right from the start,” said Dr. Tomiko Oskotsky, who directs the March of Dimes database with Dr. Sirota. “This will dramatically increase the chances of an enduring research partnership that can lead to breakthroughs in diagnostics and therapeutics for moms and babies.”
Dr. Oskotsky added that the PRC team is working to improve the beta model in ways that can benefit the research community.
“We are just getting started with ChatPTB,” she said. “It’s exciting.”