Exploring Anti-Aging Literature via ConvexTopics and Large Language Models

Area of Science:

• Biomedical Informatics
• Computational Biology
• Data Science

Background:

• The increasing volume of biomedical publications necessitates advanced methods for knowledge organization and trend identification.
• Traditional clustering and topic modeling techniques (e.g., K-means, Latent Dirichlet Allocation) often suffer from instability and local optima, hindering reproducibility.
• Existing methods lack the interpretability and scalability required for effective analysis of large biomedical datasets.

Purpose of the Study:

• To develop a novel, scalable, and interpretable clustering algorithm for biomedical text analysis.
• To address the limitations of existing methods by ensuring global optima and stable topic generation.
• To facilitate the discovery of emerging trends and the organization of knowledge within the biomedical literature.

Main Methods:

• Reformulation of a convex optimization-based clustering algorithm.
• Selection of data exemplars to ensure stable and fine-grained topic generation.
• Application to a dataset of approximately 12,000 PubMed articles on aging and longevity.

Main Results:

• The proposed method successfully uncovered interpretable topics related to aging and longevity, validated by medical experts.
• Identified topics covered diverse areas including molecular mechanisms, dietary supplements, physical activity, and gut microbiota.
• Demonstrated superior performance in terms of reproducibility and interpretability compared to K-means, LDA, and BERTopic.

Conclusions:

• The convex optimization-based clustering approach provides a robust and reproducible method for knowledge discovery in biomedical research.
• This technique enhances the organization of complex biomedical information and aids in identifying research trends.
• The developed method lays the foundation for scalable, web-accessible tools for biomedical knowledge discovery.