Background/Objectives: Many complex diseases are resulted from genetic factors and interactions. Mendelian randomization (MR) enables us to overcome problems of confounding and reverse causality. Limited MR studies consider the interaction between instrumental variables. This study was aimed to apply a Lasso for hierarchical interactions model to investigate the interaction between instrumental variables in MR. Methods: Significant genetic markers (with p-value threshold = 1 × 10−5) obtained from the genome-wide association study (GWAS) were entered into a Lasso for hierarchical interactions model to select significant interaction terms (with p-value threshold =5 × 10−2). Two-stage least square instrumental variable regression was used to evaluate the causal effects between exposures and outcomes in MR. This method was illustrated using a GWAS data of smoking cessation. The R software was used for investigating the interactions between instrumental variables in MR study. Results: A total of thirteen genetic markers were found to be associated with Fagerstrom test for nicotine dependence (FTND) scores and nine gene-gene interactions associated with FTND score were identified from the Lasso for hierarchical interactions model. After adjusting for gender, age, and duration of smoking, MR analysis showed that smokers with higher FTND scores were less likely to quit smoking at six month (OR = 0.88, 95%C.I. = [0.76, 1.03]) and at twelve month (OR = 0.93, 95%C.I. = [0.80, 1.08]), respectively. Conclusion: This study provided a novel approach for identifying interactions between instrumental variables in Mendelian randomization study.
Date:
2023-05
Relation:
European Journal of Human Genetics. 2023 May;31(Suppl. S1):297-298.
