AD-AutoGPT: An autonomous GPT for Alzheimer’s disease infodemiology

Haixing Dai, Yiwei Li, Zhengliang Liu, Lin Zhao, Zihao Wu, Suhang Song, Shen Ye, Dajiang Zhu, Xiang Li, Li Sheng, Xiaobai Yao, Lu Shi, Tai‐Quan Peng, Quanzheng Li, Zhuo Chen, Donglan Zhang, Tianming Liu, Gengchen MaiPLOS 2025 · DOI

TL;DR

This paper introduces AD-AutoGPT, an autonomous AI system designed to analyze complex health narratives, specifically focusing on Alzheimer's Disease (AD). By leveraging large language models (LLMs) and adapting the AutoGPT framework, AD-AutoGPT automates the collection, processing, and analysis of news data, offering valuable insights into public discourse surrounding AD.

What Problem Does This Paper Address?

Understanding the public discourse around complex health issues like Alzheimer's Disease is crucial for public health initiatives. Traditionally, this involves labor-intensive methods like web scraping and manual data analysis, which struggle to keep pace with the sheer volume of information available online. The authors aim to address this by developing an autonomous system that can efficiently gather, process, and analyze vast amounts of textual data related to AD, providing a more nuanced and timely understanding of public perception and research trends.

Methods

The core of this work is AD-AutoGPT, an LLM-based tool inspired by AutoGPT. It utilizes a "chain-of-thinking" approach, breaking down complex user prompts into smaller, manageable tasks. AD-AutoGPT employs a set of customized tools within an instruction library to perform various functions:

  • Data Collection: Searching and saving news articles from authoritative sources (Alzheimer's Association, BBC, Mayo Clinic, NIA).
  • Text Summarization: Using GPT-4 to condense large volumes of news text, employing a map-reduce method to handle token limits.
  • Spatiotemporal Information Extraction: Identifying and extracting geographical locations (using geoparsing tools like GeoText) and timestamps from news articles.
  • Topic Modeling: Applying Latent Dirichlet Allocation (LDA) to identify and visualize the main themes and trends in the news discourse.
  • Visualization: Generating maps of news locations and histograms of news frequency over time, as well as streamgraphs to illustrate topic trends.

The system is designed to be autonomous, requiring only user prompts to initiate and execute the entire data analysis pipeline.

Datasets

The study utilized a dataset of 277 news articles collected from four key sources: the Alzheimer's Association, BBC, Mayo Clinic, and the National Institute on Aging, covering the period from June 2022 to May 2023. This dataset was used to validate the functionalities of AD-AutoGPT.

Key Insights and Ideas

  • Autonomous Infodemiology: AD-AutoGPT demonstrates the potential of LLMs to automate complex infodemiology tasks, significantly reducing manual effort and increasing efficiency in public health research.
  • Nuanced Understanding of AD Discourse: By analyzing news trends, spatial distribution, and key topics, AD-AutoGPT provides a data-driven understanding of public focus and concerns surrounding Alzheimer's Disease.
  • Tailored LLM Application: The authors highlight specific improvements over generic AutoGPT for public health, including specialized prompting for information retrieval, enhanced spatiotemporal extraction, and more robust summarization and analysis capabilities.
  • Foundation for Future AI in Health: This work lays the groundwork for broader AI applications in public health, suggesting that similar autonomous systems could be developed for other diseases.

Key Results

  • AD-AutoGPT successfully collected and processed 277 news articles, performing text extraction, summarization, spatiotemporal analysis, and hot topic identification without human intervention.
  • Spatiotemporal Analysis: Visualizations showed a global distribution of news, with concentrations in the southeastern United States and Western Europe. Temporal analysis revealed a declining trend in news reports from June 2022 to May 2023, with peaks in September, October, and November 2022.
  • Topic Modeling: LDA analysis identified five key topics in AD news, including:
  • Drug Treatment: Focus on protein, lipid, and drug research.
  • Personal Medication: Emphasis on individual treatment, amyloid, and tissue.
  • Causes of AD: Exploring sleep, brain, blood, and cell-related aspects, reflecting popular science interest.
  • Future Outlook: Discussions on future plans and expectations for AD research.
  • Diagnosis and Care: Public concerns about diagnosis, caregiving, and prevention.

Limitations and Future Work

  • Cost of GPT-4: The study was limited to analyzing 277 articles due to the high cost of OpenAI's GPT-4 API.
  • Generalizability: While AD-AutoGPT is presented as a general framework, the authors suggest further refinement to adapt it to other dementias (e.g., prion disease, FTLD, LBD) and various brain disorders.
  • Data Sources: Future work could involve integrating AD-AutoGPT with other data sources like social media platforms and electronic health records for a more comprehensive perspective.
  • Ethical Considerations: The paper acknowledges ethical issues related to autonomous LLMs, including potential biases in training data, privacy concerns, and the risk of propagating misinformation. Future development should address these challenges.

Summary updated

Abstract

In this pioneering study, inspired by AutoGPT, the state-of-the-art open-source application based on the GPT-4 large language model, we develop a novel tool called AD-AutoGPT, which can conduct data collection, processing, and analysis about complex health narratives of Alzheimer's Disease in an autonomous manner via users' textual prompts. We collated comprehensive data from a variety of news sources, including the Alzheimer's Association, BBC, Mayo Clinic, and the National Institute on Aging since June 2022, leading to the autonomous execution of robust trend analyses, intertopic distance map visualization, and identification of salient terms pertinent to Alzheimer's Disease. This approach has yielded not only a quantifiable metric of relevant discourse but also valuable insights into public focus on Alzheimer's Disease. This application of AD-AutoGPT in public health signifies the transformative potential of AI in facilitating a data-rich understanding of complex health narratives like Alzheimer's Disease in an autonomous manner, setting the groundwork for future AI-driven investigations in global health landscapes. Code, a demo video, and other information are available at https://github.com/levyisthebest/AD-AutoGPT.
BibTeX
@inproceedings{dai2025ad,
  author = {Haixing Dai and Yiwei Li and Zhengliang Liu and Lin Zhao and Zihao Wu and Suhang Song and Shen Ye and Dajiang Zhu and Xiang Li and Li Sheng and Xiaobai Yao and Lu Shi and Tai‐Quan Peng and Quanzheng Li and Zhuo Chen and Donglan Zhang and Tianming Liu and Gengchen Mai},
  title = {{AD-AutoGPT: An autonomous GPT for Alzheimer’s disease infodemiology}},
  booktitle = {PLOS},
  year = {2025},
  doi = {10.1371/journal.pgph.0004383},
}
Figure