Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2020 Sep 11;15(9):e0238304.
doi: 10.1371/journal.pone.0238304. eCollection 2020.

Investigation of gene-gene interactions in cardiac traits and serum fatty acid levels in the LURIC Health Study

Jiayan Zhou  1   2 , Kristin Passero  1   3 , Nicole E Palmiero  1 , Bertram Müller-Myhsok  4   5   6 , Marcus E Kleber  7 , Winfried Maerz  7   8   9 , Molly A Hall  1   3
Affiliations

Investigation of gene-gene interactions in cardiac traits and serum fatty acid levels in the LURIC Health Study

Jiayan Zhou et al. PLoS One. .

Abstract

Epistasis analysis elucidates the effects of gene-gene interactions (G×G) between multiple loci for complex traits. However, the large computational demands and the high multiple testing burden impede their discoveries. Here, we illustrate the utilization of two methods, main effect filtering based on individual GWAS results and biological knowledge-based modeling through Biofilter software, to reduce the number of interactions tested among single nucleotide polymorphisms (SNPs) for 15 cardiac-related traits and 14 fatty acids. We performed interaction analyses using the two filtering methods, adjusting for age, sex, body mass index (BMI), waist-hip ratio, and the first three principal components from genetic data, among 2,824 samples from the Ludwigshafen Risk and Cardiovascular (LURIC) Health Study. Using Biofilter, one interaction nearly met Bonferroni significance: an interaction between rs7735781 in XRCC4 and rs10804247 in XRCC5 was identified for venous thrombosis with a Bonferroni-adjusted likelihood ratio test (LRT) p: 0.0627. A total of 57 interactions were identified from main effect filtering for the cardiac traits G×G (10) and fatty acids G×G (47) at Bonferroni-adjusted LRT p < 0.05. For cardiac traits, the top interaction involved SNPs rs1383819 in SNTG1 and rs1493939 (138kb from 5' of SAMD12) with Bonferroni-adjusted LRT p: 0.0228 which was significantly associated with history of arterial hypertension. For fatty acids, the top interaction between rs4839193 in KCND3 and rs10829717 in LOC107984002 with Bonferroni-adjusted LRT p: 2.28×10-5 was associated with 9-trans 12-trans octadecanoic acid, an omega-6 trans fatty acid. The model inflation factor for the interactions under different filtering methods was evaluated from the standard median and the linear regression approach. Here, we applied filtering approaches to identify numerous genetic interactions related to cardiac-related outcomes as potential targets for therapy. The approaches described offer ways to detect epistasis in the complex traits and to improve precision medicine capability.

PubMed Disclaimer

Conflict of interest statement

The authors have read the journal's policy and have the following competing interests: WZ and MEK are paid employees of The Synlab Holding Deutschland GmbH. There are no patents, products in development or marketed products associated with this research to declare. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

Figures

Fig 1
Fig 1. Design of quality control and gene-gene interaction analysis.
The phenotype quality control was performed in CLARITE software [15]. The genotype quality control and the principle component analysis (PCA) were performed in PLINK 1.90 [13]. The genome-wide association study (GWAS) and gene-gene interaction (G×G) were performed in PLATO [16]. SNP-SNP interaction models were assessed using a likelihood ratio test (LRT) between the reduced model and full model, where Var1 and Var2 are the potential SNP predictors of interest, Y is the phenotype, covs are the covariates, and ε is the error.
Fig 2
Fig 2. Quantile-quantile (Q-Q) plots of observed LRT p-values and referential G×Gs using two filtering methods.
G×G based on 1000 random selected SNPs was performed as a baseline to assess the inflation (Black). G×G under main effect filtering (Blue) and G×G under Biofilter filtering (Red) were performed and compared with the G×G based on 1000 random selected SNPs for each phenotype. Individual Q-Q plots were generated to visualize the observed LRT p-values to the expected LRT p-values for cardiac traits and fatty acids. The red line represents the ideal estimation of LRT p-values. The corresponding genomic inflation factors were calculated for showing the model inflation (S3 Table).
Fig 3
Fig 3. Comparison of model inflation based on genomic inflation factor among different approaches.
The model inflation for each interaction was calculated (S3 Table) using the median approach and regression approach for different filtering methods (Red: Biofilter filtering, Blue: Main effect filtering, and Black: Reference). The horizontal and vertical black dotted lines represent the inflation factor at 1 for different approaches. The line of equality was also plotted for viewing the estimation differences from models. Five models with comparably large inflation were labeled.
Fig 4
Fig 4. Distribution of interaction beta coefficients and LRT p-values for interactions in cardiac traits and fatty acids under different filtering methods.
The beta coefficients and LRT p-values for interactions are plotted by -log10(LRT p) against interaction beta coefficient. The significant interactions are highlighted according to the significance level of LRT p-values (blue: FDR-adjusted LRT < 0.05; red: Bonferroni-adjusted LRT p < 0.05).
Fig 5
Fig 5. Distribution of phenotype and genotype for the top interactions in cardiac traits under different filtering methods.
Individuals with cardiac trait were assigned “1” and individuals without cardiac trait were assigned as “-1”. The distribution of cardiac trait status and corresponding genotype for SNPs for the top interactions were plotted for A) rs10804247-rs7735781 for the patients with venous thrombosis or pulmonary embolism under Biofilter filtering, C) rs11642027-rs6986305 for the patients with venous thrombosis or pulmonary embolism under main effect filtering, and E) rs1383819-rs1493939 in for the patients with the history of hypertension under main effect filtering. The randomly selected 100 individuals were also plotted for each top interaction (B, D, F).
Fig 6
Fig 6. Distribution of phenotype and genotype for the top interactions in fatty acids under different filtering methods.
The distribution of fatty acid concentration and corresponding SNP genotype for SNPs for the top interactions were plotted for A) rs10989148-rs11711981 in log-transformed 9-cis 12-trans octadecanoic acid under main effect filtering, C) rs10829717-rs4839193 in log-transformed 9-trans 12-trans octadecanoic acid under main effect filtering, and E) rs2227016-rs351219 in log-transformed trans palmitoleic acid (C16:1n7t) under main effect filtering. The randomly selected 100 individuals were also plotted for each top interaction (B, D, F).
Fig 7
Fig 7. Genotype combinations between rs6986305 and rs11642027 for the participates with venous thrombosis or pulmonary embolism under main effect filtering.
The genotypes of each SNPs were plotted with the number of the participants with venous thrombosis or pulmonary embolism (cases: red) or not (controls: purple).
Fig 8
Fig 8. Genotypes combinations between rs7735781 and rs10804247 for the participates with venous thrombosis or pulmonary embolism under Biofilter filtering.
The genotypes of each SNPs were plotted with the number of the participants with venous thrombosis or pulmonary embolism (cases: red) or not (controls: purple).
Fig 9
Fig 9. Genotypes combinations between rs1383819 and rs1493939 for the participates with hypertension under main effect filtering.
The genotypes of each SNPs were plotted with the number of the participants with hypertension (cases: red) or not (controls: purple).
Fig 10
Fig 10. Mapping significant interactions at the gene level with Bonferroni adjusted p < 0.05.
The Bonferroni significant interactions with gene annotations were selected. Only the omega-6 fatty acid (C18:2n6tt) had Bonferroni significant interactions that mapped to genes. Genes GCNT1 and SCN4B were involved in multiple interactions.
Fig 11
Fig 11. Mapping significant interactions at the gene level with FDR less than 0.05.
The significant interactions with FDR less than 0.05 with appropriate gene annotations were selected. Two omega-6 fatty acids (C18:2n6ct in red and C18:2n6ct in blue) and the trans-palmitoleic acid (C16:1n7t in green) had significant interactions that mapped to genes. Multiple genes, including DNAH5 and NRG2, were involved in more than one interaction.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. Cole BS, Hall MA, Urbanowicz RJ, Gilbert-Diamond D, Moore JH. Analysis of Gene-Gene Interactions. Current protocols in human genetics. John Wiley & Sons, Ltd; 2017. pp. 1.14.1–1.14.10. 10.1002/cphg.45 - DOI - PubMed
    1. Zhao J, Zhu Y, Xiong M. Genome-wide gene-gene interaction analysis for next-generation sequencing. Eur J Hum Genet. 2016;24: 421–8. 10.1038/ejhg.2015.147 - DOI - PMC - PubMed
    1. Huang Y, Wang C, Yao Y, Zuo X, Chen S, Xu C, et al. Molecular Basis of Gene-Gene Interaction: Cyclic Cross-Regulation of Gene Expression and Post-GWAS Gene-Gene Interaction Involved in Atrial Fibrillation. PLoS Genet. 2015;11: e1005393 10.1371/journal.pgen.1005393 - DOI - PMC - PubMed
    1. Miao L, Yin R-X, Pan S-L, Yang S, Yang D-Z, Lin W-X. BCL3-PVRL2-TOMM40 SNPs, gene-gene and gene-environment interactions on dyslipidemia. Sci Rep. 2018;8 10.1038/S41598-018-24432-W - DOI - PMC - PubMed
    1. Kwon M, Leem S, Yoon J, Park T. GxGrare: gene-gene interaction analysis method for rare variants from high-throughput sequencing data. BMC Syst Biol. 2018;12: 19 10.1186/s12918-018-0543-4 - DOI - PMC - PubMed

Publication types

MeSH terms

Grants and funding

This work is supported by the USDA National Institute of Food and Agriculture and Hatch Appropriations under Project #PEN04275 and Accession #1018544. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Genotyping of the LURIC study participants was supported by the 7th Framework Program AtheroRemo (grant agreement #201668) of the European Union. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The Synlab Holding Deutschland GmbH provided support in the form of the salaries for author W.Z. and M.E.K. but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section.