Related Experiment Video
Updated: Feb 17, 2026
Identifying Dysregulated Genes Induced by Kaposi's Sarcoma-associated Herpesvirus KSHV
Published on: September 14, 2010
Kaposi's sarcoma in individuals living with HIV: comparative assessment of AI-based clinical responses using a
Sibel Altunisik Toplu1, Nihal Altunisik2, Dursun Turkmen2
1Department of Infectious Diseases and Clinical Microbiology, Inonu University Faculty of Medicine, Malatya, Turkey.
Objective:
To comparatively evaluate the clinical responses of three digital platforms (ChatGPT 5.2, DeepSeek, Consensus) in terms of responsiveness, accuracy, and clinical applicability, using a standardized set of questions on HIV-related Kaposi's sarcoma. Kaposi's sarcoma is a cutaneous malignancy; therefore, the relationship of this study to toxicology is indirect and methodological, focusing on information synthesis rather than toxicological evaluation.
Methods:
Ten clinical questions titled 'Kaposi's Sarcoma in HIV-Infected Patients - Standard Questionnaire' were administered to each platform in separate sessions, and responses were recorded. Answers were independently evaluated by three field experts (one Infectious Diseases and Clinical Microbiology specialist and two Dermatology specialists) using a four-point accuracy scale. Quantitative analyses focused on response availability and accuracy distribution across platforms. Qualitative characteristics, including citation practices, visual support, and presentation of clinical decision algorithms, were assessed descriptively.
Findings:
Chat GPT and Consensus answered all questions (10/10), whereas DeepSeek failed to generate a response to the KS-IRIS question due to a technical error (9/10). Comparison of accuracy category distributions across platforms revealed no statistically significant difference (Pearson chi-square test, p = 0.663). The median accuracy score was 1 (excellent) for all three platforms, with an interquartile range of 1-2. Qualitative analysis demonstrated that consistent citation of sources was observed only in Consensus, visual support was exclusive to ChatGPT, and structured clinical decision-making algorithms were most prominent in ChatGPT outputs.
Conclusion:
Although quantitative accuracy was comparable across platforms when assessed using a standardized Kaposi's sarcoma question set, notable differences were identified in qualitative features, including evidence presentation, visual support, and clinical decision structure. Artificial intelligence and literature-based digital platforms may support clinicians in complex conditions such as HIV related Kaposi's sarcoma. However, their outputs should be interpreted alongside current clinical guidelines and expert judgment.
