The authors estimated pairwise local genetic correlations across 16 cancers and a pan-cancer phenotype, then tested whether polygenic risk scores for one cancer predicted another. Genome-wide correlation identified 20 significant cancer pairs, while regional analysis found shared signals in 82 regions spanning 66 pairs, with five regions forming mutually correlated clusters enriched for inflammatory function. Decomposing the polygenic scores showed the five significant cross-cancer score associations were driven by pleiotropy groups and immune-related pathways.
Why it is interesting: Bears on whether a single polygenic score could ever stratify risk across multiple cancers at once, and suggests any shared component is narrow and pathway-specific.