B3GNT6-Linked Multimodal Signatures Integrate Tissue Morphology and PTM-Related Transcriptomics to Stratify Tumor

Colon adenocarcinoma (COAD) exhibits substantial molecular and microenvironmental heterogeneity, limiting the reliability and clinical application of single-omics biomarkers. Post-translational modifications (PTMs) play a pivotal role in connecting genotype to phenotype, yet prognostic models informed by PTMs seldom incorporate tissue architecture alongside transcriptomic states. This study developed a PTM-based multimodal framework for prognostic stratification and mechanistic analysis in COAD. Additionally, bulk RNA-seq data, clinicopathological information, and whole-slide H&E images from TCGA-COAD were integrated with single-cell RNA-seq (GSE132465) and spatial transcriptomics (GSE225857). Differentially expressed genes related to PTMs were identified, and pathomic features derived from H&E slides were fused with transcriptomic data via an autoencoder-based latent space. Prognosis-related latent features were then selected to establish a patient-level PTM-related multimodal risk score (PTMLS). A total of 102 PTM-related differentially expressed genes were enriched in glycan biosynthesis and ubiquitin-mediated pathways. The PTMLS effectively stratified patients into high- and low-risk groups, with significant differences in overall survival in both training (HR = 3.54, 95% CI 2.67-4.70, P = 0.001) and validation cohorts (HR = 2.68, 95% CI 1.49-4.82, P = 0.006). Low-risk tumors displayed higher immune and ESTIMATE scores, lower TIDE scores, distinct immunotherapy responder profiles, and differential predicted drug sensitivity. Cross-modal interpretation identified B3GNT6 as a key contributor linked to histopathological features. Single-cell and spatial analyses localized B3GNT6-associated epithelial states and their interactions with vascular and stromal compartments. Functional experiments further revealed that B3GNT6 expression was downregulated in tumors and inhibited colorectal cancer proliferation, migration, tumor growth, and metastasis. This PTM-informed multimodal framework integrates routine pathology and transcriptomic data, providing robust prognostic stratification, clinically relevant immune and therapeutic phenotyping, and mechanistic insight supporting B3GNT6 as a potential therapeutic target in COAD.