Trustworthy AI for disease pathogenesis and precision medicine: A disease-agnostic framework from data to deployment
Genes & Diseases
Cite this paper
Plain citation
Hao Chi, Xin Liu, Yalan Yan, Lexin Wang, Peng Luo, Lanqian Su, Lai Jiang, Xuancheng Zhou, Gang Huang, Yuxuan Jiang, Shiyang Yin, Ke Su, Haiqing Chen, Ge Zhang, Pengpeng Zhang, Hao Zhang, Bicheng Ye, Qiang Cao, Jie Liu, Vedbar Khadka, Hua Yang, Guanhu Yang, Jiayu Xu, Peng Wang, Xuelian Wang, Zhijia Xia, Xiaosong Li, Youping Deng, Zhenjiang Li, Ke Xu. Trustworthy AI for disease pathogenesis and precision medicine: A disease-agnostic framework from data to deployment. Genes & Diseases. 2026; article 102480. https://doi.org/10.1016/j.gendis.2026.102480
Author evidence note
Principal contribution
This review connects biomedical data provenance and endpoint design with model selection, causal interpretation, validation, and deployment governance. Its disease-agnostic framework includes an evidence ladder for positioning AI findings and a proposed model-and-data card for reproducible reporting. It synthesizes methods across omics, imaging/pathology, clinical records, and longitudinal data to guide trustworthy research and translation.
Evidence ladder
The review organizes support for an AI finding into association, robustness, biological coherence, causal support, perturbation evidence, and clinical utility. The ladder helps researchers identify what additional evidence is needed to move a finding toward a biological explanation or clinical application.
Source: Evidence ladder, manuscript lines 491–518, physical PDF pages 26–27.
Research use
The framework can inform the design and reporting of studies that combine biomedical data types. Its proposed model-and-data card gives researchers a place to document data provenance, model choices and validation conditions alongside the intended use.
Publication and source version
Genes & Diseases made this review available online on 18 September 2026 (article 102480). The Cardiff repository copy is the Journal Pre-proof under CC BY 4.0 and carries the same formal DOI.
Evidence scope and limitations
Author-maintained explanation of a published methodological review, based on the publisher's Journal Pre-proof in Cardiff University's repository and official bibliographic records.
- The article synthesizes methods and proposes a framework. Its data-availability statement reports no newly generated or analyzed data; the review does not report original clinical model validation.
Provenance
- Cardiff University repository full text — Publisher Journal Pre-proof; CC BY 4.0. Front matter; manuscript lines 121–133, 491–518, 666–679 and 738–740. Physical PDF pagination includes repository and publisher cover pages.
- Publisher-deposited citation metadata — Crossref metadata. Formal DOI, journal, 30-author byline, issued year, article number and version-of-record license metadata.
Source review: 2026-11-04.