Metabolic syndrome and colorectal cancer share GPAT3 and PLAU as candidate genes: Mendelian randomisation, machine learning, single-cell knockout and docking

Metabolic syndrome (MetS) associates with colorectal cancer (CRC), but the direction and linking nodes are unresolved. We integrated bidirectional Mendelian randomisation (MetS as a latent genetic liability), MetS–CRC transcriptomic intersection, machine-learning prioritisation, immune deconvolution, single-cell virtual knockout and docking. MR supported a MetS-to-CRC effect (inverse-variance-weighted OR = 1.231, 95% CI 1.052–1.439, P = 0.0093); the reverse direction was non-significant and, with only seven instruments, inconclusive. The transcriptomes shared 200 co-differentially expressed genes, 36 within a TNF/NF-κB prior set from MSigDB hallmark founder sets. Of seven hubs retained by all three algorithms, GPAT3 and PLAU were significant in both GSE10950 and TCGA/GEPIA2, GPAT3 being directionally concordant between the MetS and CRC contrasts and PLAU concordant only across tumour platforms. GPAT3 virtual knockout perturbed blood-vessel morphogenesis, extracellular-matrix organisation and myeloid activation. PLAU correlated most strongly with tumour neutrophils ( r = 0.376). Docking of three ligands against the PLAU catalytic domain predicted binding for all three, strongest for AGB (N-(1-adamantyl)-N′-(4-guanidinobenzyl)urea) and amiloride. These findings are consistent with a causal MetS-to-CRC direction (robust to MR-PRESSO outlier correction and MR-RAPS, and placed in context by multivariable MR) and nominate GPAT3 and PLAU as candidate nodes for testing.