Gene-based interaction analysis shows GABAergic genes interacting with parenting in adolescent depressive symptoms.

BACKGROUND: Most gene-environment interaction studies (G × E) have focused on single candidate genes. This approach is criticized for its expectations of large effect sizes and occurrence of spurious results. We describe an approach that accounts for the polygenic nature of most psychiatric phenotypes and reduces the risk of false-positive findings. We apply this method focusing on the role of perceived parental support, psychological control, and harsh punishment in depressive symptoms in adolescence. METHODS: Analyses were conducted on 982 adolescents of Caucasian origin (Mage (SD) = 13.78 (.94) years) genotyped for 4,947 SNPs in 263 genes, selected based on a literature survey. The Leuven Adolescent Perceived Parenting Scale (LAPPS) and the Parental Behavior Scale (PBS) were used to assess perceived parental psychological control, harsh punishment, and support. The Center for Epidemiologic Studies Depression Scale (CES-D) was the outcome. We used gene-based testing taking into account linkage disequilibrium to identify genes containing SNPs exhibiting an interaction with environmental factors yielding a p-value per single gene. Significant results at the corrected p-value of p < 1.90 × 10-4 were examined in an independent replication sample of Dutch adolescents (N = 1354). RESULTS: Two genes showed evidence for interaction with perceived support: GABRR1 (p = 4.62 × 10-5 ) and GABRR2 (p = 9.05 × 10-6 ). No genes interacted significantly with psychological control or harsh punishment. Gene-based analysis was unable to confirm the interaction of GABRR1 or GABRR2 with support in the replication sample. However, for GABRR2, but not GABRR1, the correlation of the estimates between the two datasets was significant (r (46) = .32; p = .027) and a gene-based analysis of the combined datasets supported GABRR2 × support interaction (p = 1.63 × 10-4 ). CONCLUSIONS: We present a gene-based method for gene-environment interactions in a polygenic context and show that genes interact differently with particular aspects of parenting. This accentuates the importance of polygenic approaches and the need to accurately assess environmental exposure in G × E.

Gene-environment interaction study on the polygenic risk score for neuroticism, childhood adversity, and parental bonding.

The present study examines whether neuroticism is predicted by genetic vulnerability, summarized as polygenic risk score for neuroticism (PRSN), in interaction with bullying, parental bonding, and childhood adversity. Data were derived from a general population adolescent and young adult twin cohort. The final sample consisted of 202 monozygotic and 436 dizygotic twins and 319 twin pairs. The Short Eysenck Personality questionnaire was used to measure neuroticism. PRSN was trained on the results from the Genetics of Personality Consortium (GPC) and United Kingdom Biobank (UKB) cohorts, yielding two different PRSN. Multilevel mixed-effects models were used to analyze the main and interacting associations of PRSN, childhood adversity, bullying, and parental bonding style with neuroticism. We found no evidence of gene-environment correlation. PRSN thresholds of .005 and .2 were chosen, based on GPC and UKB datasets, respectively. After correction for confounders, all the individual variables were associated with the expression of neuroticism: both PRSN from GPC and UKB, childhood adversity, maternal bonding, paternal bonding, and bullying in primary school and secondary school. However, the results indicated no evidence for gene-environment interaction in this cohort. These results suggest that genetic vulnerability on the one hand and negative life events (childhood adversity and bullying) and positive life events (optimal parental bonding) on the other represent noninteracting pathways to neuroticism.

A Comparison of Methods for Gene-Based Testing That Account for Linkage Disequilibrium.

Controlling the type I error rate while retaining sufficient power is a major concern in genome-wide association studies, which nowadays often examine more than a million single-nucleotide polymorphisms (SNPs) simultaneously. Methods such as the Bonferroni correction can lead to a considerable decrease in power due to the large number of tests conducted. Shifting the focus to higher functional structures (e.g., genes) can reduce the loss of power. This can be accomplished via the combination of p-values of SNPs that belong to the same structural unit to test their joint null hypothesis. However, standard methods for this purpose (e.g., Fisher's method) do not account for the dependence among the tests due to linkage disequilibrium (LD). In this paper, we review various adjustments to methods for combining p-values that take LD information explicitly into consideration and evaluate their performance in a simulation study based on data from the HapMap project. The results illustrate the importance of incorporating LD information into the methods for controlling the type I error rate at the desired level. Furthermore, some methods are more successful in controlling the type I error rate than others. Among them, Brown's method was the most robust technique with respect to the characteristics of the genes and outperformed the Bonferroni method in terms of power in many scenarios. Examining the genetic factors of a phenotype of interest at the gene-rather than SNP-level can provide researchers benefits in terms of the power of the study. While doing so, one should be careful to account for LD in SNPs belonging to the same gene, for which Brown's method seems the most robust technique.

Using information-theoretic approaches for model selection in meta-analysis.

Meta-regression can be used to examine the association between effect size estimates and the characteristics of the studies included in a meta-analysis using regression-type methods. By searching for those characteristics (i.e., moderators) that are related to the effect sizes, we seek to identify a model that represents the best approximation to the underlying data generating mechanism. Model selection via testing, either through a series of univariate models or a model including all moderators, is the most commonly used approach for this purpose. Here, we describe alternative model selection methods based on information criteria, multimodel inference, and relative variable importance. We demonstrate their application using an illustrative example and present results from a simulation study to compare the performance of the various model selection methods for identifying the true model across a wide variety of conditions. Whether information-theoretic approaches can also be used not only in combination with maximum likelihood (ML) but also restricted maximum likelihood (REML) estimation was also examined. The results indicate that the conventional methods for model selection may be outperformed by information-theoretic approaches. The latter are more often among the set of best methods across all of the conditions simulated and can have higher probabilities for identifying the true model under particular scenarios. Moreover, their performance based on REML estimation was either very similar to that from ML estimation or at times even better depending on how exactly the REML likelihood was computed. These results suggest that alternative model selection methods should be more widely applied in meta-regression.