Using ChatGPT to Predict Cancer Predisposition Genes: A Promising Tool for Pediatric Oncologists.
| Autor: | Sultan I; Department of Pediatrics, King Hussein Cancer Center, Amman, JOR., Al-Abdallat H; Department of Medicine, University of Jordan, Amman, JOR., Alnajjar Z; Department of Medicine, Hashemite University, Zarqa, JOR., Ismail L; Department of Medicine, University of Jordan, Amman, JOR., Abukhashabeh R; Department of Cell Therapy and Applied Genomics, King Hussein Cancer Center, Amman, JOR., Bitar L; Department of Pediatric Oncology, King Hussein Cancer Center, Amman, JOR., Abu Shanap M; Department of Pediatric Oncology, King Hussein Cancer Center, Amman, JOR. |
|---|---|
| Jazyk: | angličtina |
| Zdroj: | Cureus [Cureus] 2023 Oct 24; Vol. 15 (10), pp. e47594. Date of Electronic Publication: 2023 Oct 24 (Print Publication: 2023). |
| DOI: | 10.7759/cureus.47594 |
| Abstrakt: | Background: Determining genetic susceptibility for cancer predisposition syndromes (CPS) through cancer predisposition genes (CPGs) testing is critical in facilitating appropriate prevention and surveillance strategies. This study investigates the use of ChatGPT, a large language model, in predicting CPGs using clinical notes. Methods: Our study involved 53 patients with pathogenic CPG mutations. Two kinds of clinical notes were used: the first visit note, containing a thorough history and physical exam, and the genetic clinic note, summarizing the patient's diagnosis and family history. We asked ChatGPT to recommend CPS genes based on these notes and compared these predictions with previously identified mutations. Results: Rb1 was the most frequently mutated gene in our cohort (34%), followed by NF1 (9.4%), TP53 (5.7%), and VHL (5.7%). Out of 53 patients, 30 had genetic clinic notes of a median length of 54 words. ChatGPT correctly predicted the gene in 93% of these cases. However, it failed to predict EPCAM and VHL genes in specific patients. For the first visit notes (median length: 461 words), ChatGPT correctly predicted the gene in 64% of these cases. Conclusion: ChatGPT shows promise in predicting CPGs from clinical notes, particularly genetic clinic notes. This approach may be useful in enhancing CPG testing, especially in areas lacking genetic testing resources. With further training, there is a possibility for ChatGPT to improve its predictive potential and expand its clinical applicability. However, additional research is needed to explore the full potential and applicability of ChatGPT. Competing Interests: The authors have declared that no competing interests exist. (Copyright © 2023, Sultan et al.) |
| Databáze: | MEDLINE |
| Externí odkaz: |
|