Preoperative liquid biopsy transcriptomic panel for risk assessment of lymph node metastasis in T1 gastric cancer

Ping’an Ding, Jiaxiang Wu, Haotian Wu, Wenqian Ma, Tongkun Li, Peigang Yang, Honghai Guo, Yuan Tian, Jiaxuan Yang, Limian Er, Renjun Gu, Lilong Zhang, Ning Meng, Xiaolong Li, Zhenjiang Guo, Lingjiao Meng, Qun ZhaoJournal of Experimental & Clinical Cancer Research 2025 · DOI

TL;DR

Researchers have developed a non-invasive liquid biopsy test that uses a 4-mRNA transcriptomic panel to predict lymph node metastasis (LNM) in early-stage (T1) gastric cancer. By combining these molecular biomarkers with clinical data, their "RSA model" could prevent more than half of the unnecessary radical surgeries currently performed on patients misclassified as high-risk.

What Problem Does This Paper Address?

Early-stage (T1) gastric cancer is increasingly detectable thanks to better screening. For many of these patients, a minimally invasive endoscopic procedure (ESD) is enough to cure the cancer—if the cancer hasn't spread to the lymph nodes.

The catch is that current clinical tools, like the "eCura" scoring system, are notoriously imprecise. To avoid missing a single case of metastasis, these systems cast a very wide net, classifying roughly 70–80% of T1 patients as "high risk." In reality, only about 8–16% of these patients actually have lymph node metastasis (LNM). This massive gap leads to "overtreatment," where thousands of patients undergo radical gastrectomies (stomach removal) and extensive lymph node dissections that they didn't actually need, exposing them to surgical complications and lifelong digestive issues.

Methods

The authors employed a multi-stage discovery and validation workflow to find a better way to predict LNM:

  1. Biomarker Discovery: They cross-referenced transcriptomic data from public databases (TCGA and GEO) with in-house mRNA sequencing of T1 gastric cancer tissues. They were looking for genes that were consistently upregulated in patients with LNM compared to those without.
  2. The 4-mRNA Panel: They identified a specific signature consisting of four genes: SDS, TESMIN, NEB, and GRB14.
  3. Model Construction: They developed a Risk Stratification Assessment (RSA) model. This isn't just a genetic test; it’s a composite score that combines the expression levels of those four mRNAs (measured via RT-qPCR) with three clinical red flags: tumor size, depth of infiltration, and lymphovascular invasion.
  4. Cross-Platform Validation: Crucially, they tested whether this panel worked across different sample types: fresh-frozen surgical tissue, small gastroscopic biopsies, and—most importantly—peripheral blood (liquid biopsy).

Datasets

The study relied on a massive amount of data across several cohorts:

  • Discovery Sets: TCGA (412 cancer, 36 normal tissues) and GEO GSE246963 (8 LNM+, 8 LNM-).
  • Tissue Validation: 404 fresh-frozen specimens from six Chinese diagnostic centers (218 training, 186 validation).
  • Biopsy Validation: 122 matched gastroscopic biopsy specimens.
  • Liquid Biopsy Validation: 315 serum samples (147 training, 168 validation).
  • Prospective Cohort: 97 patients (ChiCTR-IIR-17011197) with blood samples taken both before and three months after surgery.
  • Pan-Cancer Specificity: Samples from esophageal (n=38), colon (n=32), and rectal (n=29) cancers to test if the panel was specific to the stomach.

Key Insights and Ideas

  • From Tissue to Blood: The most significant "aha" moment was that the 4-mRNA signature identified in tumor tissue translated effectively to a liquid biopsy. This means a simple blood draw could potentially replace or supplement invasive pathological staging.
  • The Specificity Factor: The panel isn't just a general "cancer marker." When tested against other gastrointestinal cancers (esophageal, colorectal), the 4-mRNA panel was significantly more accurate for gastric cancer (AUC 0.879 vs ~0.6-0.7 for others).
  • Biological "Why": Using single-cell sequencing and pathway analysis, the authors found these four genes are linked to tumor progression pathways like angiogenesis and PI3K/AKT/mTOR signaling, and are expressed in the immune microenvironment (T cells and B cells), suggesting they reflect the tumor's ability to "prime" the lymph nodes for spread.
  • Post-Op Drop: In the prospective cohort, the expression of these four mRNAs dropped significantly after the tumor was surgically removed, confirming that the signals measured in the blood were indeed coming from the primary gastric tumor.

Key Results

The RSA model consistently outperformed traditional clinical assessments:

  • Predictive Accuracy: The model achieved an Area Under the Curve (AUC) of 0.890 in tissue and 0.873 in liquid biopsy training cohorts. In the independent liquid biopsy validation set, it maintained a strong AUC of 0.852.
  • Reducing Overtreatment: This is the headline result. Using traditional criteria, the overtreatment rate was 84.4%. The liquid-biopsy RSA model slashed that rate to 56.0%. In the tissue-based validation, it was even more dramatic, dropping overtreatment from 83.9% to 44.1%.
  • Clinical Utility: The model showed a "superior net benefit" across a wide range of threshold probabilities, meaning it is robust enough for real-world clinical decision-making.

Limitations and Future Work

While promising, the authors acknowledge several hurdles:

  • Retrospective Bias: Most of the data came from retrospective cohorts, which can introduce selection bias.
  • Demographics: The study was conducted entirely on Chinese patients. Because gastric cancer genetics and clinical characteristics can vary by ethnicity, the model needs validation in Western and other Asian populations.
  • Sample Size: While the total N is large, the number of patients who actually had LNM (the positive cases) was relatively small, which can affect the stability of the model.
  • Long-term Outcomes: The study lacks long-term survival data to prove that patients spared from surgery by this model actually have the same survival rates as those who undergo radical gastrectomy.

The next step? Large-scale, international prospective clinical trials to move this from the lab to the clinic.

Summary updated

Abstract

We developed and validated a novel liquid biopsy-based RSA model that accurately predicts LNM in T1 GC patients. This non-invasive approach could significantly reduce unnecessary surgical interventions and optimize treatment strategies for high-risk T1 GC patients.
BibTeX
@article{ding2025preoperative,
  author = {Ping’an Ding and Jiaxiang Wu and Haotian Wu and Wenqian Ma and Tongkun Li and Peigang Yang and Honghai Guo and Yuan Tian and Jiaxuan Yang and Limian Er and Renjun Gu and Lilong Zhang and Ning Meng and Xiaolong Li and Zhenjiang Guo and Lingjiao Meng and Qun Zhao},
  title = {{Preoperative liquid biopsy transcriptomic panel for risk assessment of lymph node metastasis in T1 gastric cancer}},
  journal = {Journal of Experimental \& Clinical Cancer Research},
  year = {2025},
  doi = {10.1186/s13046-025-03305-x},
}
Figure