Scholarlink

Purpose <p>Blood-based biomarkers for the early detection of gastric cancer remain limited. We conducted a systematic analysis of pre-treatment plasma proteomics profiled using the Olink Reveal panel and developed diagnostic prediction models within a leakage-controlled, outer out-of-fold (OOF) internal validation framework based on nested cross-validation. In parallel, pathway-level activity was summarized using a predefined exploratory suppressive burden index to support biological interpretation.</p> Methods <p>We analysed plasma samples from 90 participants (30 healthy controls, 30 early-stage/localized gastric cancer, 30 advanced gastric cancer) using the Olink Reveal platform. Group E comprises stage T1 and T2 tumours; consequently, this group is operationally defined as early-stage/localized gastric cancer, rather than classic early gastric cancer in the strict sense. The advanced group comprised T3/T4 disease. Differential abundance was tested with age- and sex-adjusted limma models. Predictive models were developed using nested cross-validation, with preprocessing and tuning performed within training folds and performance evaluated through outer-loop OOF predicted probabilities. The suppressive burden index used predefined Reactome terms and explicit orientation signs.</p> Results <p>Age- and sex-adjusted analyses identified several proteins differentially abundant between early gastric cancer and healthy controls. Reactome term-set analyses indicated immune-related pathway variation across disease states, although these pathway-level findings should be interpreted as exploratory. OOF internal validation showed that proteomic models improved discrimination over age/sex covariates for the main early-detection task, while the relative performance of protein-only and full models varied across resampling analyses. Calibration slopes and decision-curve results were reported from OOF predictions and interpreted as internal-validation evidence only.</p> Conclusion <p>Plasma proteomics captures systemic immune alterations associated with early-stage/localized gastric cancer and may support minimally invasive risk modeling. Given the modest single-center cohort and absence of external validation, the diagnostic models and suppressive burden index should be considered exploratory and require independent validation before clinical use.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Exploration of plasma proteomic landscapes using Olink Reveal technology to identify biomarkers for early-stage/localized gastric cancer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月29日

查看全文
  • Shao rui Ding,
  • Xiang Li,
  • Xiang Dong,
  • Ping Chen

摘要

Purpose

Blood-based biomarkers for the early detection of gastric cancer remain limited. We conducted a systematic analysis of pre-treatment plasma proteomics profiled using the Olink Reveal panel and developed diagnostic prediction models within a leakage-controlled, outer out-of-fold (OOF) internal validation framework based on nested cross-validation. In parallel, pathway-level activity was summarized using a predefined exploratory suppressive burden index to support biological interpretation.

Methods

We analysed plasma samples from 90 participants (30 healthy controls, 30 early-stage/localized gastric cancer, 30 advanced gastric cancer) using the Olink Reveal platform. Group E comprises stage T1 and T2 tumours; consequently, this group is operationally defined as early-stage/localized gastric cancer, rather than classic early gastric cancer in the strict sense. The advanced group comprised T3/T4 disease. Differential abundance was tested with age- and sex-adjusted limma models. Predictive models were developed using nested cross-validation, with preprocessing and tuning performed within training folds and performance evaluated through outer-loop OOF predicted probabilities. The suppressive burden index used predefined Reactome terms and explicit orientation signs.

Results

Age- and sex-adjusted analyses identified several proteins differentially abundant between early gastric cancer and healthy controls. Reactome term-set analyses indicated immune-related pathway variation across disease states, although these pathway-level findings should be interpreted as exploratory. OOF internal validation showed that proteomic models improved discrimination over age/sex covariates for the main early-detection task, while the relative performance of protein-only and full models varied across resampling analyses. Calibration slopes and decision-curve results were reported from OOF predictions and interpreted as internal-validation evidence only.

Conclusion

Plasma proteomics captures systemic immune alterations associated with early-stage/localized gastric cancer and may support minimally invasive risk modeling. Given the modest single-center cohort and absence of external validation, the diagnostic models and suppressive burden index should be considered exploratory and require independent validation before clinical use.