Elamur Research⌕
☰
home / papers / 2026 / BioTopicXplor: A Web Tool for Interactive Exploration of PubMed Literature through Reproducible Topics.

BioTopicXplor: A Web Tool for Interactive Exploration of PubMed Literature through Reproducible Topics.

Paper · 2026 · Lana Yeganova, Donald Comeau, Won Kim et al.

Summary

PubMed finds articles for a query but does not show how a result set is structured conceptually. Companion tools such as Anne O'Tate, BioTextQuest v2.0 and GoPubMed organize literature through ontologies, user-chosen terms or features. Clustering methods like K-means can also give different topics on each run. The authors present BioTopicXplor, a free web tool that takes a PubMed query and retrieves titles and abstracts, by default from the past five years. It then discovers subtopics with ConvexTopics. This method treats clustering as a convex problem whose candidate centers are exemplars taken from the data. As a result, it reaches a global optimum, needs no preset topic count and returns the same subtopics on repeated runs. It operates on enriched single terms and noun phrases compared with the Dice coefficient. GPT-5.1 writes PMID-linked summaries and titles from each subtopic's ten top documents and groups subtopics into major topics, which users can browse or search. A case study on 12,000 anti-aging articles illustrates the workflow. The tool's value lies in pairing stable, data-driven topics with LLM output anchored to them, which limits hallucination.

Key points

  • Cluster handling: Clusters that share more than 50% of their documents are removed. Each subtopic gets three scores: intrinsic quality, size, and visibility (how prominent its top-10 terms are in its top-10 documents).
  • Evaluation:
    • ConvexTopics was evaluated in a companion paper on 20-Newsgroups, Reuters-21578 and PubMed sets (Anti-Aging, Diabetes Mellitus, Age-Related Macular Degeneration). It compared favorably with K-means, LDA and BERTopic.
    • Major topics were checked only by manual inspection and may vary somewhat between runs.
  • Runtime: For the query "familial Mediterranean fever", 1,099 documents took 65 minutes and 5,731 documents took 87 minutes. LLM summarization and grouping, not clustering, account for most of the time.
  • Implementation: The front end uses HTML, CSS and JavaScript, and the back end uses C++ and Python on Linux. It runs on a shared 12-core Xeon Gold machine (up to 2.30 GHz) with 196 GB RAM.
  • Anti-aging use case: The analysis indicated that the safety evidence for widely sold NAD supplements is weak. Gaps include bioavailability, long-term safety studies and optimal dosing.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🏛️National Institutes of HealthBethesda · United States13 works · 3,846 citations🔬United States National Library of MedicineBethesda · United States8 works · 5,342 citations

 Fields, topics and concepts

📚 Shelves 8The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksComputer Vision and Pattern Recognition · 3,486 worksSocial Sciences · 1,954 worksBiochemistry, Genetics and Molecular Biology · 263 worksMolecular Biology · 260 worksGeneral Social Sciences · 254 works

Cite

Cite as Lana Yeganova, Donald Comeau, Won Kim, Natalie Xie, Shubo Tian, W John Wilbur and Zhiyong Lu. 2026. BioTopicXplor: A Web Tool for Interactive Exploration of PubMed Literature through Reproducible Topics. BioNLP 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.bionlp-1.37

@inproceedings{yeganova2026biotopicxplor,
  title = {{BioTopicXplor: A Web Tool for Interactive Exploration of PubMed Literature through Reproducible Topics.}},
  author = {Lana Yeganova, Donald Comeau, Won Kim, Natalie Xie, Shubo Tian, W John Wilbur and Zhiyong Lu},
  booktitle = {BioNLP 2026},
  year = {2026},
  doi = {10.18653/v1/2026.bionlp-1.37},
  url = {https://doi.org/10.18653/v1/2026.bionlp-1.37}
}

References

9 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Keith D Engwall. 2017. Anne o’tate. Journal of the Medical Library Association: JMLA, 105(2):200. find
  2. Stuart Lloyd. 1982. Least squares quantization in pcm. IEEE transactions on information theory, 28(2):129– 137. find
  3. Qin Song, Xiaofeng Zhou, Kexin Xu, Sishi Liu, Xinqiang Zhu, and Jun Yang. 2023. The safety and antiaging effects of nicotinamide mononucleotide in human clinical trials: an update. Advances in nutrition, 14(6):1416–1435. find
  4. Theodosios Theodosiou, Konstantinos Vrettos, Ismini Baltsavia, Fotis Baltoumas, Nikolas Papanikolaou, Andreas N Antonakis, Dimitrios Mossialos, Christos A Ouzounis, Vasilis J Promponas, Makrina Karaglani, and 1 others. 2024. Biotextquest v2. 0: an evolved tool for biomedical literature mining and co… find
  5. Danial Lashkari and Polina Golland. 2007. Convex clustering with exemplar-based models. Advances in neural information processing systems, 20. find
  6. Lana Yeganova, Donald C Comeau, Won Kim, and W John Wilbur. 2009. How to interpret pubmed queries and why it matters. Journal of the American Society for Information Science and Technology, 60(2):264–274. find
  7. David D. Lewis, Yiming Yang, Tony G. Rose, and Fan Li. 2004. Rcv1: A new benchmark collection for text categorization research. Journal of Machine Learning Research, 5:361–397.
  8. Lana Yeganova, Donald C Comeau, and W John Wilbur. 2011. Machine learning with naturally labeled data for identifying abbreviation definitions. BMC bioinformatics, 12(Suppl 3):S6. find
  9. Lana E. Yeganova, Won G. Kim, Shubo Tian, Natalie Xie, Donald C. Comeau, W. John Wilbur, and Zhiyong Lu. 2026. Exploring anti-aging literature via convextopics and large language models. AMIA Amplify Informatics Conference. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yeganova-2026-biotopicxplor-web-tool-interactive-exploration/