Abstract: 

Objective Metabolic biomarkers are expected to decode the phenotype of gastric cancer (GC) and lead to high-performance blood tests towards GC diagnosis and prognosis. We attempted to develop diagnostic and prognostic models for GC based on plasma metabolic information.

Design We conducted a large-scale, multicentre study comprising 1944 participants from 7 centres in retrospective cohort and 264 participants in prospective cohort. Discovery and verification phases of diagnostic and prognostic models were conducted in retrospective cohort through machine learning and Cox regression of plasma metabolic fingerprints (PMFs) obtained by nanoparticle-enhanced laser desorption/ionisation-mass spectrometry (NPELDI-MS). Furthermore, the developed diagnostic model was validated in prospective cohort by both NPELDI-MS and ultra-performance liquid chromatography-MS (UPLC-MS).

Results We demonstrated the high throughput, desirable reproducibility and limited centre-specific effects of PMFs obtained through NPELDI-MS. In retrospective cohort, we achieved diagnostic performance with areas under curves (AUCs) of 0.862–0.988 in the discovery (n=1157 from 5 centres) and independent external verification dataset (n=787 from another 2 centres), through 5 different machine learning of PMFs, including neural network, ridge regression, lasso regression, support vector machine and random forest. Further, a metabolic panel consisting of 21 metabolites was constructed and identified for GC diagnosis with AUCs of 0.921–0.971 and 0.907–0.940 in the discovery and verification dataset, respectively. In the prospective study (n=264 from lead centre), both NPELDI-MS and UPLC-MS were applied to detect and validate the metabolic panel, and the diagnostic AUCs were 0.855–0.918 and 0.856–0.916, respectively. Moreover, we constructed a prognosis scoring system for GC in retrospective cohort, which can effectively predict the survival of GC patients.

Conclusion We developed and validated diagnostic and prognostic models for GC, which also contribute to advanced metabolic analysis towards diseases, including but not limited to GC.

Figure 1 Schematics of PMFs for gastric cancer (GC) diagnosis and prognosis. (A) Enrolment and overview of study cohort. The retrospective cohort was consisted of 1944 subjects enrolled from 1 November 2007 to 31 August 2019 in 7 centres across China, while the prospective cohort included 264 subjects from 1 January 2021 to 31 December 2021 in lead centre. (B) Microarrayed nanoparticle-enhanced laser desorption/ionisation mass spectrometry (NPELDI-MS) was employed to construct PMFs, and ultra-performance liquid chromatography-MS (UPLC-MS) was applied in prospective cohort for validation. (C) Machine learning of PMFs was conducted to diagnose the GC patients, and Cox regression was applied to build a prognosis prediction model based on PMFs.

Figure 2 Construction of GC-associated plasma metabolic fingerprints (PMFs). (A) Scanning electron microscope image of ferric nanoparticles with designed nanoscale surface roughness. (B) Elemental mappings of the nanoparticle–glucose hybrids are shown with C in yellow, Fe in green and O in red, respectively, illustrating the trapping of metabolites by significant higher carbon signal on the nanoparticles as compared with the background (p<0.05). The scale bars represent 200 nm. (C) Digital image shows the microarrays after printing samples and nanoparticles, and the scale bar represents that the distance between two adjacent points is 5 mm. (D) Typical mass spectra of plasma samples from non-GC group (upper) and GC group (bottom). (E) Heatmap of PMFs for 1944 plasma samples collected from seven centres, containing 300 m/z signals for each sample through data preprocessing. The colour scale was processed by logarithmic correction. (F) Frequency distribution of similarity scores in non-GC group (upper) and GC group (bottom), and over 95% of samples shared similarity scores over 0.9 in both two groups. (G) The unsupervised t-distributed stochastic neighbour embedding (t-SNE) map, (H) principal component analysis (PCA) score plot and (I) uniform manifold approximation and projection (UMAP) visualisation of PMFs of 1944 samples with colours annotated by centres, demonstrating limited centre-specific effects. GC, gastric cancer.

Figure 3 Machine learning in PMFs towards gastric cancer (GC) diagnosis. Scores generated by machine learning, taking neural network (NN) as an example, of individuals of each fold in discovery dataset (A) and independent external verification dataset (B). The scores of GC patients (red) are significantly higher than non-GC participants (blue, p<0.05). The receiver operating characteristic curves generated by NN, ridge regression (RR), lasso regression (LR), support vector machine (SVM) and random forest (RF) with area under curves (AUCs) of 0.909–0.988 in discovery dataset (C) and AUCs of 0.871–0.979 in independent external verification dataset (D). The precision-recall curves (PRC) generated by NN, RR, LR, SVM and RF with areas under PRC (AUPRCs) of 0.895–0.983 in discovery dataset (E) and AUPRCs of 0.874–0.978 in independent external verification dataset (F).

DOI：10.1136/gutjnl-2023-330045