Federal medical scientists are turning to large language models to handle an explosion of research data, with agencies like the National Cancer Institute deploying AI tools to streamline workflows and accelerate discoveries.
AI tools help researchers sift through mounting data
The volume of medical research has surged in recent years. Cancer-related publications in the National Library of Medicine’s PubMed database, for example, have more than doubled since 2005. To manage this deluge, federal agencies are integrating large language models into their operations.
Related: Healthcare Leaders Face New AI Bill of Materials
The National Cancer Institute launched NanCI—Connecting Scientists—in 2024, a mobile and web application that summarizes key research findings and recommends relevant papers based on users’ interests. The tool draws from sources like PubMed and Semantic Scholar, offering biomedical researchers a way to quickly identify trends and potential collaborators.
“NanCI is one of our most visible applications of large language models,” said Nastaran Zahir, acting director of the Center for Cancer Training at NCI. “We’re looking at ways AI can support different aspects of the research and training ecosystem, whether it can help with knowledge discovery or administrative efficiency.”
Related: HRSA revives 340B rebate pilot program
The app’s Ask NanCI feature allows researchers to interact with scientific content conversationally. Users can ask questions like, “Tell me the major outcome of this paper,” and the system generates responses using the paper’s content and LLM capabilities. Google Cloud provides the underlying infrastructure for the tool, which is accessible on iOS, Android, and web browsers.
Accuracy and validation remain key challenges
Despite their potential, LLMs come with risks, including hallucinations and inaccurate outputs. Agencies have issued guidance on managing these risks, emphasizing the need for continuous evaluation. “There’s a lot of testing around how more autonomy can lead to increased productivity in different areas—evidence synthesis, biomedical text extraction, finding patients for clinical trials,” said Wes Anderson, a quantitative medicine scientist at the Critical Path Institute. “But as these processes become more automated, we need more validation.”
Related: Hospital Groups Urge Becerra to Delay Info‑Blocking Rule
The goal isn’t to replace scientists but to augment their work. “We aren’t trying to use these models to replace scientists,” Anderson said. “We’re trying to augment some of the more tedious parts of our workflows.”
NCI’s Zahir sees NanCI as a step toward an “AI-native scientific infrastructure,” where tools actively support research rather than just storing information. “It represents a shift toward an AI-native scientific infrastructure, where AI tools can actively support how science is done,” she said. For now, the focus remains on refining these systems to ensure they deliver reliable, actionable insights—without introducing new bottlenecks.