Pan-cancer study detects genetic risk variants and shared genetic basis in two large cohorts

Abstract

Deciphering the shared genetic basis of distinct cancers has the potential to elucidate carcinogenic mechanisms and inform broadly applicable risk assessment efforts. Here, we undertake genome-wide association studies (GWAS) and comprehensive evaluations of heritability and pleiotropy across 18 cancer types in two large, population-based cohorts: the UK Biobank (408,786 European ancestry individuals; 48,961 cancer cases) and the Kaiser Permanente Genetic Epidemiology Research on Adult Health and Aging cohorts (66,526 European ancestry individuals; 16,001 cancer cases). The GWAS detect 21 genome-wide significant associations independent of previously reported results. Investigations of pleiotropy identify 12 cancer pairs exhibiting either positive or negative genetic correlations; 25 pleiotropic loci; and 100 independent pleiotropic variants, many of which are regulatory elements and/or influence cross-tissue gene expression. Our findings demonstrate widespread pleiotropy and offer further insight into the complex genetic architecture of cross-cancer susceptibility.

Fig. 1. Cross-cancer genetic correlations (r_g) calculated via LD-score regression (LDSC) and associated cancers from the locus-specific pleiotropy analysis.

a Cancer pairs are connected if the genetic correlation had P < 0.05, width of the line is proportional to magnitude of r_g, color of the line indicates direction of correlation (red is negative and blue is positive), and shading is proportional to strength of association according to P, where the Bonferroni-corrected threshold is 0.05/153 = 3.27 × 10⁻⁴; b cancer pairs are connected by a line (each line represents one region) if a region contains any SNPs associated with either cancer, where regions are formed around index SNPs with P < 5 × 10⁻⁸ for any cancer in the cancer-specific meta-analyses and SNPs are added if they have P < 5 × 10⁻⁸ for any cancer, are within 500 kb of the index SNP, and have LD r² > 0.5 with the index SNP.

Fig. 2. Manhattan plot displaying one-directional variant-specific pleiotropy from ASSET.

The red dashed line represents the genome-wide significance threshold (P < 5 × 10⁻⁸), and the black dotted line represents a suggestive threshold (P < 1 × 10⁻⁶). Highlighted in purple are genome-wide significant loci where the overall pleiotropic P is less than all individual P for the selected cancers. Highlighted in green are the genome-wide significant loci where the overall pleiotropic P is greater than at least one of the individual P for the selected cancers. All highlighted loci are independent of bidirectional SNPs with smaller overall P.

Fig. 3. Manhattan plot displaying bidirectional variant-specific pleiotropy from ASSET.

The red dashed line represents the genome-wide significance threshold (P < 5 × 10⁻⁸), and the black dotted line represents a suggestive threshold (P < 1 × 10⁻⁶). Highlighted are loci with overall pleiotropic P < 5 × 10⁻⁸, the two directional P < 0.05, and not in LD with a one-directional SNP with smaller P. Loci in purple are genome-wide significant loci where the overall pleiotropic P is less than all individual P for the selected cancers, and loci in green are genome-wide significant loci where the overall pleiotropic P is greater than at least one of the individual P for the selected cancers.

Fig. 4. Summary of cancer pairs associated with and functional consequences of the 100 one- and bidirectional pleiotropic variants.

a The number of pleiotropic variants (of the independent 100 one- and bidirectional variants with overall pleiotropic P < 5 × 10⁻⁸) associated with each pair of cancers by type of pleiotropic effect for select cancer pairs using ASSET: SNPs identified in the one-directional analysis, where all associations are in the same direction (navy); SNPs identified in the bidirectional analysis, where both cancers in the pair are associated in the same direction (both risk increasing or both risk decreasing), even though at least one other cancer is associated in the opposite direction (blue); and SNPs identified in the bidirectional analysis, where the pair of cancers are associated in opposite directions (one risk increasing and one risk decreasing) (green). b The distribution of variant consequences and corresponding enrichment, calculated using Fisher’s exact test comparing the proportion of variants belonging to each functional class observed among the 100 ASSET variants to all variants in the UK Biobank. Pleiotropic variants were enriched in intergenic (P = 0.043) and non-coding RNA transcripts (P = 0.015). c Venn diagram summarizing the number of variants with specific regulatory elements, based on analyses of chromatin features from Roadmap and expression quantitative trait loci (eQTL) associations. d Distribution of DeepSEA functional significance scores, providing an integrated summary score based on evolutionary conservation and chromatin data, with 0 denoting variants most likely to be functional.