Machine learning model in multi-omics perspective demystifies the prognostic significance of crotonylation heterogeneity in clear cell renal cell carcinoma.

Background: Crotonylation, a post-translational modification, is implicated in cancer progression, but its prognostic significance in clear cell renal cell carcinoma (ccRCC) remains unclear. This study aimed to demystify crotonylation heterogeneity and establish a robust prognostic model for ccRCC.

Methods: Using multi-omics approaches, we analyzed transcriptomic data from TCGA-KIRC and GEO cohorts (GSE40435, GSE167573, GSE29609). Crotonylation scores were calculated via ssGSEA, with related gene modules identified through WGCNA. We integrated 10 machine learning algorithms to develop a prognostic model. Immune microenvironment was profiled using Cibersort, mutation landscapes via maftools, and drug sensitivity through oncoPredict. Spatial transcriptomics and single-cell data were analyzed for expression patterns, validated by qRT-PCR in 786-O and HK-2 cell lines.

Results: Dysregulation of 16/18 crotonylation-related genes was observed in ccRCC. WGCNA revealed crotonylation related modules significantly enriched in angiogenesis, calcium/Ras signaling, and cancer stemness pathways. A 5-gene prognostic model (PLCL1, DNASE1L3, CD248, CDH13, PDGFD) demonstrated robust stratification: High-risk patients showed poorer overall survival, higher Treg infiltration, elevated tumor mutation burden and increased sensitivity to several chemotherapy approaches like Cisplatin. Molecular docking identified diacetylmorphine as a potential therapeutic agent (binding energy: -7.278 kcal/mol with DNASE1L3). Spatial/single-cell analyses confirmed cell-type-specific gene expression and the diffferential expression between tumor and normal cell lines was validated by qRT-PCR.

Conclusion: This study establishes a crotonylation-based prognostic model that effectively stratifies ccRCC risk and elucidates key mechanisms linking crotonylation heterogeneity to immune evasion, mutational burden, and metabolic reprogramming. The model offers clinical utility for personalized therapy selection.