AI and Cancer Research ── Gastric ── 2026-11-09
Gastric cancer: treatment, trials and AI
- Abstract
BACKGROUND: Gastric cancer (GC) exhibits substantial molecular heterogeneity and a complex tumor microenvironment (TME). Apoptotic protease-activating factor 1 (APAF1) is a core component of mitochondrial apoptosis; however, its association with immune characteristics, spatial heterogeneity and compound discovery in GC remains poorly defined. METHODS: The bulk transcriptomics, single-cell RNA sequencing and spatial transcriptomics with machine learning-based modeling, immune deconvolution and spatial niche analysis were integrated to characterize APAF1-associated features in GC. Machine learning-guided computational screening combining molecular docking and molecular dynamics (MD) simulations was used to prioritize morellic acid (MA). The biological effects of MA and the involvement of APAF1-associated signaling were evaluated through in vitro and in vivo assays using APAF1-targeting siRNA and shRNA, respectively. RESULTS: APAF1 expression was increased in GC, particularly in early-stage tumors and was independently associated with improved overall survival after multivariable adjustment (HR = 0.681, 95% CI: 0.494-0.940, P = 0.0196). However, its standalone prognostic discrimination was limited, with time-dependent ROC analyses showing poor predictive performance. Tumor-versus-normal ROC analysis demonstrated differential expression between GC and normal tissues (AUC = 0.875, 95% CI: 0.823-0.918). Single-cell and spatial analyses revealed that APAF1 expression was associated with distinct immune-cell patterns and tumor spatial niches. Machine learning-guided computational screening prioritized MA as a candidate compound. MA inhibited GC cell proliferation and migration (IC50: 1.807 ± 0.16 to 6.216 ± 0.23 μM) and suppressed tumor growth in vivo, achieving tumor inhibition rates of 79.32 ± 7.69%, 92.37 ± 4.21%, and 92.19 ± 2.99% at low, medium and high doses, respectively. APAF1 silencing attenuated MA-associated changes in cleaved Caspase-3 activation and GSDME-NT accumulation, suggesting the involvement of APAF1-associated cell-death signaling. CONCLUSION: APAF1 is associated with distinct tumor microenvironment characteristics and clinical features in GC, supporting its potential relevance to tumor biology beyond its canonical role in mitochondrial apoptosis. The novelty of this study lies in extending the investigation of APAF1 beyond its canonical apoptotic function by integrating machine learning with multimodal transcriptomic and spatial analyses to characterize APAF1-associated TME features and computationally prioritize MA. All the findings provide a cross-scale framework linking APAF1-associated tumor biology with candidate compound screening and MA-related cell-death responses in GC.
Journal IF-equivalent: 10.4 (OpenAlex 2-year mean citedness, value as of 2026-11-08, retrieved 2026-11-09; not the official Clarivate IF)Reference: Jia B, Yan B, Li S, Yang J, Cheng Z, Zhou X, et al. Machine learning-guided multimodal analysis of APAF1-associated tumor microenvironment features prioritizes morellic acid for gastric cancer. Phytomedicine. 2026 Nov;162:158895. doi:10.1016/j.phymed.2026.158895.Checked: Abstract only