Lung cancer remains the leading cause of cancer-related mortality worldwide, with early detection critical for improving survival rates, yet conventional methods like CT scans often yield high false-positive rates. This study introduces M-GNN, a graph neural network framework leveraging GraphSAGE, to enhance early lung cancer detection through metabolomics. We constructed a heterogeneous graph integrating metabolomics data from 800 plasma samples (586 cases, 214 controls) with demographic features and Human Metabolome Database annotations, employing GraphSAGE and GAT layers for inductive learning on 107 metabolites, pathways, and diseases. M-GNN achieved a test accuracy of 89% and an ROC-AUC of 0.92, with rapid convergence within 400 epochs and robust performance across ten random seeds; key predictors included age, height, choline, Valine, Betaine, and Fumaric Acid, reflecting smoking and metabolic dysregulation. This framework offers a scalable, interpretable tool for precision oncology, surpassing benchmarks by capturing complex biological interactions, though limitations like synthetic data biases and computational demands suggest future validation with real-world cohorts and optimization. M-GNN advances lung cancer screening, promising improved survival through early detection and personalized strategies.
Lung cancer remains the leading cause of cancer-related mortality worldwide, with early detection critical for improving survival rates, yet conventional methods like CT scans often yield high false-positive rates. This study introduces M-GNN, a Graph Neural Network framework leveraging GraphSAGE, to enhance early lung cancer detection through metabolomics. We constructed a heterogeneous graph integrating metabolomics data from 800 plasma samples (586 cases, 214 controls) with demographic features and Human Metabolome Database annotations, employing GraphSAGE and GAT layers for inductive learning on 107 metabolites, pathways, and diseases. M-GNN achieved a test accuracy of 93% and an ROC-AUC of 0.96, with rapid convergence within 400 epochs and robust performance across ten random seeds; key predictors included cigarette pack years, choline, and taurine, reflecting smoking and metabolic dysregulation. This framework offers a scalable, interpretable tool for precision oncology, surpassing benchmarks by capturing complex biological interactions, though limitations like synthetic data biases and computational demands suggest future validation with real-world cohorts and optimization. M-GNN advances lung cancer screening, promising improved survival through early detection and personalized strategies.
Plasma metabolomics offers a promising avenue for the non-invasive classification of lung cancer. However, most classification frameworks treat lung cancer as a single entity, conflating non-small-cell lung cancer (NSCLC) with pulmonary neuroendocrine neoplasms (NENs), despite their distinct biology and treatment pathways. We apply a heterogeneous graph neural network to a three-class problem discriminating Control, NSCLC, and NEN from targeted plasma metabolomic profiling in 800 participants (466 NSCLC, 120 NEN, 214 Controls). Matched metabolites were annotated by Cell Danger Response (CDR) relevance class and expected direction of change, with these annotations encoded as metabolite features and direction-aware patient–metabolite edge weights. Across ten random seeds, the three-class model achieved a macro-F1 of and macro-AUROC of . Collapsing the three-class posteriors to a cancer-versus-control score yielded an AUROC of , a sensitivity of , and an F1 score of . A separate cancer-only classifier distinguished NEN from NSCLC with an accuracy of and an AUROC of . The gradient-times-input attribution identified one-carbon, glycine–serine, proline, and ornithine–arginine metabolic programs, with proline, C5DC, uric acid, and fumaric acid among the leading annotated metabolites. Removal of all 11 cohort-derived Class N annotations left NSCLC-versus-NEN AUROC essentially unchanged, and broader comparator analyses showed no measurable predictive advantage of the CDR prior over the otherwise matched concentration-weighted GNN. These findings indicate that predictive discrimination is driven primarily by the measured metabolomic features, whereas the CDR component provides a biologically structured framework for directional metabolite and pathway interpretation. Independent external and prospective validation is required to establish generalizability and clinical utility.