EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

A pilot evaluation of ChatGPT-4o-assisted hypothesis generation for exploring potential links between fungal biology and cervical cancer

Yu Li, Tianshu Gu, Li Ma, Lan Yao, Minghui Li, Shuyan Han, Xingjiang Li, Weikuan Gu, Jiamin Ma

Antonie van Leeuwenhoek · 2026

DOI: 10.1007/s10482-026-02434-3 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Antonie van Leeuwenhoek, 2026. Verfasst von Yu Li und 8 weiteren Autor:innen. Beteiligte Einrichtung: Department of Pathology, Qiqihar Medical University, No. 333 Bukui North Street, Qiqihar, 161006, Heilongjiang, China. Hauptthemen laut MEDLINE: Uterine Cervical Neoplasms, Fungi. Die Arbeit nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: Qiqihar City Joint Guidance Project of Science and Technology Program, Health Science Center, University of Tennessee, the Second Affiliated Hospital of Zhengzhou University.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The potential relationship between fungal biology and cervical cancer remains poorly understood. This study aimed to assess the ability of ChatGPT-4o to generate research hypotheses concerning possible links between fungal factors, HPV persistence, and cervical carcinogenesis. ChatGPT-4o generated hypotheses in three prespecified domains: microbial dysbiosis, inflammation, and immune modulation. Three author-evaluators with expertise in pathology, oncology, and microbiology independently assigned overall evaluation scores and assessed the scientific reasonableness of experimental-plan components. Targeted literature searches identified publications directly addressing each proposed mechanism. Inter-rater reliability was quantified using a two-way random-effects, absolute-agreement intraclass correlation coefficient (ICC) and pairwise quadratic-weighted Cohen's kappa. No human-generated hypothesis set, alternative language model, or repeated independent AI session was included. ChatGPT-4o generated 55 hypotheses. No directly matching publication was identified for 48 hypotheses under the prespecified search framework; this finding was interpreted as indicating an underexplored direction, not proof of novelty. Based on the predefined composite evaluation rubric, 35 hypotheses (63.6%) received high mean evaluation scores and 20 (36.4%) received moderate mean evaluation scores. Agreement for hypothesis scores was high (ICC (2,1) = 0.975; pairwise weighted kappa = 0.963-1.000). Among 18 experimental-plan components, 12 received high mean scores and 6 received moderate mean scores; agreement was lower but still strong (ICC (2,1) = 0.821; weighted kappa = 0.727-0.914). Representative hypotheses concerned fungal metabolites and oxidative stress, fungal-associated inflammation and oxidative damage, and fungal modulation of immune-checkpoint pathways. This pilot study suggests that ChatGPT-4o can generate diverse and potentially testable hypotheses concerning fungal biology and cervical cancer. However, the generated hypotheses should be regarded as exploratory research directions rather than validated mechanisms and require independent literature verification and experimental validation.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Uterine Cervical NeoplasmsFungiFemalePilot ProjectsHumansGenerative Artificial IntelligenceLarge Language ModelsReproducibility of Results

Stichwörter der Autor:innen

Artificial intelligenceCervical cancerChatGPTFungiHypothesis

Autor:innen und Institutionen

  • Department of Pathology, Qiqihar Medical University, No. 333 Bukui North Street, Qiqihar, 161006, Heilongjiang, China.
  • Department of Clinical Pharmacy and Translational Science, University of Tennessee Health Science Center, 920 Madison Avenue, Memphis, 38103, TN, USA.
  • Department of Orthopedic Surgery and BME-Campbell Clinic, University of Tennessee Health Science Centre, 1211 Union Ave, Memphis, 38163, TN, USA.
  • Center of Integrative Research, The First Hospital of Qiqihar City, 30 Gongyuan Road, Qiqihar, 161005, Heilongjiang, China.
  • College of Health Management, Harbin Medical University, Heilongjiang 157 Baojian Road, Harbin, 150081, Heilongjiang, China.
  • Key Laboratory of Carcinogenesis and Translational Research (Ministry of Education), Department of Integration of Chinese and Western Medicine, Peking University Cancer Hospital and Institute, No. 52 Fucheng Road, Beiji…

Förderung

  • Qiqihar City Joint Guidance Project of Science and Technology Program
  • Health Science Center, University of Tennessee
  • the Second Affiliated Hospital of Zhengzhou University

Zitieren

APA

Li, Y., Gu, T., Ma, L., Yao, L., Li, M., Han, S., Li, X., Gu, W., & Ma, J. (2026). A pilot evaluation of ChatGPT-4o-assisted hypothesis generation for exploring potential links between fungal biology and cervical cancer. Antonie van Leeuwenhoek. https://doi.org/10.1007/s10482-026-02434-3

Vancouver

Li Y, Gu T, Ma L, Yao L, Li M, Han S, et al. A pilot evaluation of ChatGPT-4o-assisted hypothesis generation for exploring potential links between fungal biology and cervical cancer. Antonie van Leeuwenhoek. 2026. doi:10.1007/s10482-026-02434-3

BibTeX
@article{Li2026,
  author = {Li, Yu and Gu, Tianshu and Ma, Li and Yao, Lan and Li, Minghui and Han, Shuyan and Li, Xingjiang and Gu, Weikuan and Ma, Jiamin},
  title = {A pilot evaluation of ChatGPT-4o-assisted hypothesis generation for exploring potential links between fungal biology and cervical cancer},
  journal = {Antonie van Leeuwenhoek},
  year = {2026},
  doi = {10.1007/s10482-026-02434-3}
}
Kostenlos starten →