Metabolomic machine learning predictor for diagnosis and prognosis of gastric cancer

Yangzi Chen,Bohong Wang,Yizi Zhao,Xinxin Shao,Mingshuo Wang,Fuhai Ma,Laishou Yang,Meng Nie,Peng Jin,Ke Yao,Haibin Song,Shenghan Lou,Hang Wang,Tianshu Yang,Yantao Tian,Peng Han,Zeping Hu

DOI: https://doi.org/10.1038/s41467-024-46043-y

IF: 16.6

2024-02-24

Nature Communications

Abstract:Gastric cancer (GC) represents a significant burden of cancer-related mortality worldwide, underscoring an urgent need for the development of early detection strategies and precise postoperative interventions. However, the identification of non-invasive biomarkers for early diagnosis and patient risk stratification remains underexplored. Here, we conduct a targeted metabolomics analysis of 702 plasma samples from multi-center participants to elucidate the GC metabolic reprogramming. Our machine learning analysis reveals a 10-metabolite GC diagnostic model, which is validated in an external test set with a sensitivity of 0.905, outperforming conventional methods leveraging cancer protein markers (sensitivity < 0.40). Additionally, our machine learning-derived prognostic model demonstrates superior performance to traditional models utilizing clinical parameters and effectively stratifies patients into different risk groups to guide precision interventions. Collectively, our findings reveal the metabolic landscape of GC and identify two distinct biomarker panels that enable early detection and prognosis prediction respectively, thus facilitating precision medicine in GC.

multidisciplinary sciences

What problem does this paper attempt to address?

The main issue this paper attempts to address is the early diagnosis and prognosis prediction of Gastric Cancer (GC). Specifically, the researchers aim to develop a non-invasive biomarker panel through metabolomics analysis and machine learning methods to achieve early detection of gastric cancer and risk stratification of patients.

### Main Issues:
1. **Early Diagnosis**:
- Gastric cancer is a highly lethal cancer, and early diagnosis is crucial for improving clinical outcomes. However, current gold standard diagnostic methods (such as endoscopy) are invasive and costly, limiting their widespread clinical application.
- Therefore, there is an urgent need to develop a highly sensitive and specific non-invasive detection method to facilitate the early detection of gastric cancer.

2. **Prognosis Prediction**:
- Currently, clinical prognosis prediction mainly relies on the experience of surgeons, based on clinical indicators such as tumor location, TNM staging information, and pathological characteristics, which have limited accuracy.
- Therefore, developing a more accurate method to predict patient prognosis and stratify patients into different risk groups for precise intervention is a pressing issue.

### Solutions:
1. **Metabolomics Analysis**:
- Researchers conducted targeted metabolomics analysis on 702 plasma samples from multiple centers, revealing metabolic reprogramming in gastric cancer.
- Through machine learning analysis, the researchers developed a diagnostic model containing 10 metabolites (10-DM model), which showed high sensitivity (0.905) in external test sets, outperforming traditional cancer protein biomarker-based methods (sensitivity <0.40).

2. **Prognosis Model**:
- The researchers also developed a prognosis model containing 28 metabolites (28-PM model) using machine learning methods, which performed excellently in predicting patient prognosis, with a C-index of 0.83, surpassing traditional models based on clinical parameters.
- This model can effectively stratify patients into different risk groups, guiding precision medicine.

### Summary:
Through metabolomics analysis and machine learning methods, the researchers successfully developed a non-invasive biomarker panel that not only enables early diagnosis of gastric cancer but also accurately predicts patient prognosis, providing new tools and methods for precision medicine in gastric cancer.