Genomic structural equation modelling of 12 cancers identifies a latent pan-cancer susceptibility factor and shared determinants of carcinogenesis

Background Common cancers share germline susceptibility, yet the genetic architecture of a truly pan-cancer component of carcinogenesis remains incompletely resolved. Methods We applied genomic structural equation modelling (gSEM) to genome-wide association studies (GWAS) summary statistics for 12 cancers (>600,000 European-ancestry cases) to separate a latent pan-cancer genetic factor from cancer-specific components, and then performed a multivariate GWAS of that factor. We conducted comprehensive downstream genetic, predictive, and functional analyses to characterise the resulting pan-cancer factor. Findings We identified 133 LD-independent loci associated with the pan-cancer factor, including eight loci not previously associated with any cancer phenotype. Pan-cancer genetic risk was concentrated in evolutionarily conserved regulatory DNA, with enrichment in promoter and enhancer annotations and tissue-relevant signals in mammary and uterine contexts. A pan-cancer polygenic risk score (PRS) derived from the latent-factor GWAS improved prediction of overall cancer in UK Biobank (Royston-Sauerbrei R 2 = 2.97%), showed significant predictive value for most assessed cancers (R 2 up to 8.66%), and retained predictive signal in an independent East Asian validation dataset (Nagelkerke R 2 = 0.56%), outperforming a previous meta-analysis-based cross-cancer PRS and all site-specific PRSs. Gene and pathway prioritisation identified 132 putative risk genes across 106 loci, including 21 genes with little or no prior cancer annotation, and implicated canonical genome maintenance and cell cycle programmes alongside less emphasised processes involving organelle organisation, vesicle trafficking, and protein post-translational modification. Finally, proteome-wide Mendelian randomisation identified 23 blood proteins with putative causal effects on pan-cancer risk, including six druggable targets. Interpretation Together, these results delineate a gSEM-derived pan-cancer genetic architecture, provide a cross-site PRS for overall cancer susceptibility, and nominate genes and circulating proteins for functional follow-up and prevention-oriented target discovery. Funding Noncommunicable Chronic Diseases-National Science and Technology Major Project, National Natural Science Foundation of China, and National Key Research and Development Program of China.