IDDF2024-ABS-0229 Early diagnosis and prognosis prediction of gastric cancer by high-performance serum lipidome fingerprints: a multi-cohort study

<h3>Background</h3> Early detection is warranted to improve the prognosis of gastric cancer (GC), but it remains challenging. Liquid biopsy combined with machine learning will provide new insights into the diagnostic strategies of GC. Lipid metabolism reprogramming plays a crucial role in the initiation and development of tumors. Therefore, we aimed to map the lipid metabolism landscape of GC and develop novel lipid-based tools of liquid biopsy for early diagnosis and risk stratification of GC. <h3>Methods</h3> This cohort study enrolled 944 participants, including 440 patients with GC, 71 patients with gastric precancerous lesions and 433 healthy donors (IDDF2024-ABS-0229 Figure 1. Study design). Based on the lipid detection of the serums from participants by ultra-performance liquid chromatography-mass spectrometry, the serum lipid metabolic landscape of GC was depicted. Linear discriminant analysis and clustering analysis were employed to establish and validate the serum lipid metabolic signature (SLMS) and the gastric cancer prognostic subtype (GCPS), which were applied for the diagnosis and prognosis prediction of GC, respectively. Besides, multi-omics analysis was performed to further interpret the lipid metabolism disturbance in GC tissues and demonstrate the reliability of SLMS and GCPS. <h3>Results</h3> The SLMS, consisting of 19 lipids (IDDF2024-ABS-0229 Figure 2. 19 lipids in the serum lipid metabolic signature), showed excellent performances in discriminating between GC patients and healthy donors (training cohort, AUC=0.993; testing cohort, AUC=0.989; external validation cohort, AUC=0.965; predictive cohort, AUC=0.978; (IDDF2024-ABS-0229 Figure 3. The diagnostic performance of SLMS in identifying GC)). Importantly, the SLMS could efficiently distinguish early-stage GC (stage I/II) patients from healthy donors (training cohort, AUC=0.992; testing cohort, AUC=0.996; external cohort, AUC=0.975; predictive cohort, AUC=0.963; (IDDF2024-ABS-0229 Figure 4. The efficacy of SLMS in diagnosing early SATGE GE)). Moreover, the GCPS effectively predicted high-risk GC patients with poor overall survival, especially patients with advanced GC (stage III/IV, (IDDF2024-ABS-0229 Figure 5. The GCPS can predict the survival of GC patients)). Finally, the alteration of lipid metabolism in GC and the metabolic heterogeneity between two prognostic subtypes were verified in situ while some potential gene targets for GCPS were identified (IDDF2024-ABS-0229 Figure 6. Multi-omics revealed the lipid metabolism disturbance in GC). <h3>Conclusions</h3> Lipid metabolic reprogramming arose along with the development of gastric cancer. We could efficiently screen GC patients and predict their survival through the SLMS and GCPS.