Dataset Information

A fully adjusted two-stage procedure for rank-normalization in genetic association studies.

ABSTRACT: When testing genotype-phenotype associations using linear regression, departure of the trait distribution from normality can impact both Type I error rate control and statistical power, with worse consequences for rarer variants. Because genotypes are expected to have small effects (if any) investigators now routinely use a two-stage method, in which they first regress the trait on covariates, obtain residuals, rank-normalize them, and then use the rank-normalized residuals in association analysis with the genotypes. Potential confounding signals are assumed to be removed at the first stage, so in practice, no further adjustment is done in the second stage. Here, we show that this widely used approach can lead to tests with undesirable statistical properties, due to both combination of a mis-specified mean-variance relationship and remaining covariate associations between the rank-normalized residuals and genotypes. We demonstrate these properties theoretically, and also in applications to genome-wide and whole-genome sequencing association studies. We further propose and evaluate an alternative fully adjusted two-stage approach that adjusts for covariates both when residuals are obtained and in the subsequent association test. This method can reduce excess Type I errors and improve statistical power.

SUBMITTER: Sofer T

PROVIDER: S-EPMC6416071 | biostudies-literature | 2019 Apr

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A fully adjusted two-stage procedure for rank-normalization in genetic association studies.

Sofer Tamar T Zheng Xiuwen X Gogarten Stephanie M SM Laurie Cecelia A CA Grinde Kelsey K Shaffer John R JR Shungin Dmitry D O'Connell Jeffrey R JR Durazo-Arvizo Ramon A RA Raffield Laura L Lange Leslie L Musani Solomon S Vasan Ramachandran S RS Cupples L Adrienne LA Reiner Alexander P AP Laurie Cathy C CC Rice Kenneth M KM

Genetic epidemiology 20190117 3

When testing genotype-phenotype associations using linear regression, departure of the trait distribution from normality can impact both Type I error rate control and statistical power, with worse consequences for rarer variants. Because genotypes are expected to have small effects (if any) investigators now routinely use a two-stage method, in which they first regress the trait on covariates, obtain residuals, rank-normalize them, and then use the rank-normalized residuals in association analys ...[more]

PMID: 30653739

Similar Datasets

Project description:In the area of genetic epidemiology, studies of the genotype-phenotype associations have made significant contributions to human complicated trait genetics. These studies depend on specialized statistical methods for uncover the association between traits and genetic variants, both common and rare variants. Often, in analyzing such studies, potentially confounding factors, such as social and environmental conditions, are required to be involved. Multiple linear regression is the most widely used type of regression analysis when the outcome of interest is quantitative traits. Many statistical tests for identifying genotype-phenotype associations using linear regression rely on the assumption that the traits (or the residuals) of the regression follow a normal distribution. In genomic research, the rank-based inverse normal transformation (INT) is one of the most popular approaches to reach normally distributed traits (or normally distributed residuals). Many researchers believe that applying the INT to the non-normality of the traits (or the non-normality of the residuals) is required for valid inference, because the phenotypic (or residual) outliers and non-normality have the significant influence on both the type I error rate control and statistical power, especially under the situation in rare-variant association testing procedures. Here we propose a test for exploring the association of the rare variant with the quantitative trait by using a fully adjusted full-stage INT. Using simulations we show that the fully adjusted full-stage INT is more appropriate than the existing INT methods, such as the fully adjusted two-stage INT and the INT-based omnibus test, in testing genotype-phenotype associations with rare variants, especially when genotypes are uncorrelated with covariates. The fully adjusted full-stage INT retains the advantages of the fully adjusted two-stage INT and ameliorates the problems of the fully adjusted two-stage INT for analysis of rare variants under non-normality of the trait. We also present theoretical results on these desirable properties. In addition, the two available methods with non-normal traits, the quantile/median regression method and the Yeo-Johnson power transformation, are also included in simulations for comparison with these desirable properties.

Project description:BackgroundAnchored covariate-adjusted indirect comparisons inform reimbursement decisions where there are no head-to-head trials between the treatments of interest, there is a common comparator arm shared by the studies, and there are patient-level data limitations. Matching-adjusted indirect comparison (MAIC), based on propensity score weighting, is the most widely used covariate-adjusted indirect comparison method in health technology assessment. MAIC has poor precision and is inefficient when the effective sample size after weighting is small.MethodsA modular extension to MAIC, termed two-stage matching-adjusted indirect comparison (2SMAIC), is proposed. This uses two parametric models. One estimates the treatment assignment mechanism in the study with individual patient data (IPD), the other estimates the trial assignment mechanism. The first model produces inverse probability weights that are combined with the odds weights produced by the second model. The resulting weights seek to balance covariates between treatment arms and across studies. A simulation study provides proof-of-principle in an indirect comparison performed across two randomized trials. Nevertheless, 2SMAIC can be applied in situations where the IPD trial is observational, by including potential confounders in the treatment assignment model. The simulation study also explores the use of weight truncation in combination with MAIC for the first time.ResultsDespite enforcing randomization and knowing the true treatment assignment mechanism in the IPD trial, 2SMAIC yields improved precision and efficiency with respect to MAIC in all scenarios, while maintaining similarly low levels of bias. The two-stage approach is effective when sample sizes in the IPD trial are low, as it controls for chance imbalances in prognostic baseline covariates between study arms. It is not as effective when overlap between the trials' target populations is poor and the extremity of the weights is high. In these scenarios, truncation leads to substantial precision and efficiency gains but induces considerable bias. The combination of a two-stage approach with truncation produces the highest precision and efficiency improvements.ConclusionsTwo-stage approaches to MAIC can increase precision and efficiency with respect to the standard approach by adjusting for empirical imbalances in prognostic covariates in the IPD trial. Further modules could be incorporated for additional variance reduction or to account for missingness and non-compliance in the IPD trial.

Project description:As genome-wide association studies expand beyond populations of European ancestry, the role of admixture will become increasingly important in the continued discovery and fine-mapping of variation influencing complex traits. Although admixture is commonly viewed as a confounding influence in association studies, approaches such as admixture mapping have demonstrated its ability to highlight disease susceptibility regions of the genome. In this study, we illustrate a powerful two-stage testing strategy designed to uncover trait-associated single nucleotide polymorphisms in the presence of ancestral allele frequency differentiation. In the first stage, we conduct an association scan by using predicted genotypic values based on regional admixture estimates. We then select a subset of promising markers for inclusion in a second-stage analysis, where association is tested between the observed genotype and the phenotype conditional on the predicted genotype. We prove that, under the null hypothesis, the test statistics used in each stage are orthogonal and asymptotically independent. Using simulated data designed to mimic African-American populations in the case of a quantitative trait, we show that our two-stage procedure maintains appropriate control of the family wise error rate and has higher power under realistic effect sizes than the one-stage testing procedure in which all markers are tested for association simultaneously with control of admixture. We apply the proposed procedure to a study of height in 201 African-Americans genotyped at 108 ancestry informative markers. The two-stage procedure identified two statistically significant markers rs1985080 (PTHB1/BBS9) and rs952718 (ABCA12). PTHB1/BBS9 is downregulated by parathyroid hormone in osteoblastic cells and is thought to be involved in parathyroid hormone action in bones and may play a role in height. ABCA12 is a member of the superfamily of ATP-binding cassette transporters and its potential involvement in height is unclear.

Dataset Information

A fully adjusted two-stage procedure for rank-normalization in genetic association studies.

Publications

A fully adjusted two-stage procedure for rank-normalization in genetic association studies.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets