Fig 1.
Pathway-level genetic interaction models.
(A) Between-pathway interaction and between-pathway model. Two biological pathways share a common function necessary for maintaining a healthy state. Genetic variants in individual pathways do not result in a phenotype, but joint mutations in both pathways in the same individual results in disease. Between-pathway interactions clustering between two complementary pathways and appear are referred to as an instance of the between-pathway model (BPM). (B) Within-pathway interaction and within-pathway model. A single pathway supports a function for maintaining a healthy state. A single genetic variant does not result in a phenotype, but joint mutations in the same pathway results in the loss of function and a disease state. Within-pathway interactions clustered within the single pathway are called a within-pathway model (WPM). (C) Overview of the framework for discovering pathway-level genetic interactions from GWAS breast cancer data, leveraging the BridGE method [23].
Table 1.
Information about the 4 GWAS data sets used in this study.
Fig 2.
An example between pathway interaction identified from the BPC3 cohort.
(A) Interaction between Acute Myeloid Leukemia (AML) gene set and Steroid hormone biosynthesis (SHB) gene set. White and yellow nodes represent the SNPs mapped to genes in the corresponding pathways and their color shows the significance of a univariate test in the same breast cancer cohort (white: not significant; yellow: marginally significant, 10−4 < p < 0.05). Red lines indicate the risk associated SNP-SNP interactions between SNPs mapped to the corresponding pathways. (B) Null distribution of the SNP-SNP interaction density between the AML and SHB based on 200,000 SNP permutations. The arrow indicates the observed SNP-SNP interaction density in the BPC3 cohort. (C) Distribution of the significance of pairwise SNP-SNP interactions (-log10 p-value) tested individually for SNP pairs supporting the AML-SHB interaction. The most significant SNP-SNP interaction results in an FDR = 0.94 after multiple hypothesis correction, suggesting that there is not sufficient power to detect SNP-SNP interactions between these pathways in this cohort.
Fig 3.
Summary of between-pathway and within-pathway interactions discovered from the phs000812 (BPC3) cohort.
(A) Network representation of a set of significant (FDR ≤ 0.25) pathway-level interactions (BPM, WPM and PATH) that are associated with increased risk of breast cancer. Node size reflects the interaction degree. (B & C) Heatmap view of the interaction between the Acute Myeloid Leukemia (AML) gene set and the steroid hormone biosynthesis (SHB) pathway in the discovery cohort BPC3 (B) and in the replication cohort CGEMS (C). Red in the heatmap indicates that there is at least one SNP-SNP interaction identified between the corresponding genes.
Table 2.
Summary of discoveries across five breast cancer cohorts.
Fig 4.
Consensus summary of pathway-level interactions discovered from the 6 GWAS breast cancer cohorts.
(A) Network view of the most significant between-pathway interactions (BPM) (geometric mean p ≤ 5.0 × 10−5) that are supported by at least two cohorts. The supporting cohorts are indicated by the edge labels. (B) List of all within-pathway interactions (WPM) and hub pathways (PATH) that are most significant (geometric mean p ≤ 5.0 × 10−3) and supported by at least two cohorts.
Fig 5.
Network view of the between-pathway interactions (BPM) from the consensus analysis.
All BPMs satisfying a geometric mean p ≤ 5.0 × 10−4 threshold from consensus analysis are plotted. Red edges indicate interactions associated with increased breast cancer risk while green edges indicate interactions associated with decreased risk. Node size is proportional to the number of BPMs connected to each pathway. Several of the highly connected pathways are labeled by numbers, and their corresponding pathway names are listed. The complete information for these pathways can be found in S5 Table.
Fig 6.
Consensus interactions with the glutathione conjugation pathway.
(A) Gene interaction degree (fold enrichment) of all glutathione conjugation genes in the three cohorts that support a PATH interaction for the glutathione conjugation pathway (LAT517, CHN799 and JPN517). (B) Between pathway interactions associated with glutathione conjugation that are significant (FDR ≤ 0.25) in both LAT517 and CHN799 datasets. The red edges indicate they are all associated with increased risk of breast cancer. (C) Detailed statistics for the between pathway interactions shown in (B).