{
  "abstract": "Background Gastric cancer (GC) remains a major cause of cancer-related mortality, particularly in China, where delayed diagnosis contributes to poor outcomes. Current diagnostic approaches are limited by suboptimal sensitivity, invasiveness, or cost. Circulating small non-coding RNAs (sncRNAs) are promising non-invasive biomarkers because of their stability in blood and disease-associated expression patterns. However, most studies have focused on a single sncRNA class, especially microRNAs (miRNAs), and the diagnostic value of integrating multiple sncRNA types in GC remains unclear. This study aimed to systematically characterize circulating sncRNAs in serum from patients with GC and develop a robust multi-modal sncRNA-based model for non-invasive GC diagnosis.Methods Participants were recruited from the First Affiliated Hospital of Wenzhou Medical University and enrolled in discovery and independent validation phases, each including 30 patients with GC and 30 age- and sex-matched healthy controls. Five sncRNA classes, including microRNAs, PIWI-interacting RNAs (piRNAs), small nuclear RNAs (snRNAs), small nucleolar RNAs (snoRNAs), and transfer RNA-derived small RNAs (tsRNAs), were profiled by high-throughput sequencing. Candidate biomarkers were selected based on statistical and classification performance. An ensemble learning framework integrating five machine learning algorithms was constructed in the discovery cohort and evaluated in the validation cohort. A final model was established using data from both cohorts.Results A multi-modal sncRNA signature with strong diagnostic value for GC was identified. Compared with single-class sncRNA models, the integrated model achieved superior classification performance, and the ensemble framework outperformed individual machine learning algorithms. In the independent validation cohort, the multi-modal model achieved an area under the receiver operating characteristic curve (AUC) of 0.960 (95% confidence interval [CI]: 0.917-1.000), with sensitivity of 93.3% (95% CI: 83.9%-100.0%), specificity of 89.3% (95% CI: 77.3%-100.0%), and accuracy of 91.4% (95% CI: 84.5%-98.3%). After integration of both cohorts, the final 10-sncRNA model achieved an AUC of 0.970.Conclusions We established a multi-modal circulating sncRNA-based ensemble model for non-invasive GC diagnosis. Integrating multiple sncRNA classes improved diagnostic performance beyond single-class approaches, supporting the value of multi-modal sncRNA signatures as blood-based biomarkers for GC detection. Further work will include qPCR validation and large-scale multicenter prospective studies to confirm the clinical utility and generalizability of this model.",
  "authors": [
    {
      "affiliations": [
        "Department of Surgery, The Chinese University of Hong Kong, Hong Kong"
      ],
      "name": "Yijuan Wang"
    },
    {
      "affiliations": [
        "Department of Oncology, The First Affiliated Hospital of Wenzhou Medical University, China"
      ],
      "name": "Zhexin Zhang"
    },
    {
      "affiliations": [
        "Department of Oncology, The First Affiliated Hospital of Wenzhou Medical University, China"
      ],
      "name": "Jingxuan Xu"
    },
    {
      "affiliations": [
        "Department of Surgery, The Chinese University of Hong Kong, Hong Kong"
      ],
      "name": "Hua Xue"
    },
    {
      "affiliations": [
        "Department of Surgery, The Chinese University of Hong Kong, Hong Kong"
      ],
      "name": "Shuqi Chen"
    },
    {
      "affiliations": [
        "Department of Surgery, The Chinese University of Hong Kong, Hong Kong"
      ],
      "name": "Hang Jiang"
    },
    {
      "affiliations": [
        "Department of Oncology, The First Affiliated Hospital of Wenzhou Medical University, China"
      ],
      "name": "Jinfei Chen"
    },
    {
      "affiliations": [
        "Department of Surgery, The Chinese University of Hong Kong, Hong Kong"
      ],
      "name": "Xin Wang"
    }
  ],
  "title": "IDDF2026-ABS-0147 Multi-modal small non-coding RNA biomarker model for non-invasive diagnosis of gastric cancer",
  "uid": "0bdb5ac7-6df4-5212-8cf7-3d696e9cfda5"
}