1. Deriving stratified effects from joint models investigating Gene-Environment Interactions
- Author
-
Mary F. Feitosa, DC Rao, Timothy D. Majarian, Vincent Laville, Hugues Aschard, Yun Ju Sung, Paul S. de Vries, Amy R. Bentley, Alisa K. Manning, Département de Biologie Computationnelle - Department of Computational Biology, Institut Pasteur [Paris] (IP)-Centre National de la Recherche Scientifique (CNRS), Broad Institute of MIT and Harvard (BROAD INSTITUTE), Harvard Medical School [Boston] (HMS)-Massachusetts Institute of Technology (MIT)-Massachusetts General Hospital [Boston], The University of Texas Health Science Center at Houston (UTHealth), National Institutes of Health [Bethesda] (NIH), Washington University School of Medicine [Saint Louis, MO], Harvard T.H. Chan School of Public Health, This work was supported by the HL118305 grant from the NHLBI. HA was also supported by R21HG007687 from NHGRI. PSdV was supported by American Heart Association grant number 18CDA34110116. This research was supported in part by the Intramural Research Program of the National Human Genome Research Institute in the Center for Research in Genomics and Global Health (CRGGH—Z01HG200362). CRGGH is also supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), Center for Information Technology, and the Office of the Director at the National Institutes of Health., Institut Pasteur [Paris]-Centre National de la Recherche Scientifique (CNRS), and Washington University School of Medecine [Saint Louis, MO]
- Subjects
Computer science ,Marginal model ,lcsh:Computer applications to medicine. Medical informatics ,Interaction ,Biochemistry ,03 medical and health sciences ,0302 clinical medicine ,Structural Biology ,Dummy variable ,Statistics ,Humans ,Binary exposure ,MESH: Models, Genetic ,Gene–environment interaction ,lcsh:QH301-705.5 ,Molecular Biology ,030304 developmental biology ,[STAT.AP]Statistics [stat]/Applications [stat.AP] ,0303 health sciences ,MESH: Humans ,Models, Genetic ,Applied Mathematics ,MESH: Gene-Environment Interaction ,Estimator ,Environmental exposure ,Stratified analysis ,Summary statistics ,Gene-environment interaction ,Computer Science Applications ,Term (time) ,MESH: Joints ,lcsh:Biology (General) ,[SDV.GEN.GH]Life Sciences [q-bio]/Genetics/Human genetics ,Sample size determination ,030220 oncology & carcinogenesis ,lcsh:R858-859.7 ,Joints ,[SDV.SPEE]Life Sciences [q-bio]/Santé publique et épidémiologie ,[INFO.INFO-BI]Computer Science [cs]/Bioinformatics [q-bio.QM] ,[STAT.ME]Statistics [stat]/Methodology [stat.ME] ,Software ,030217 neurology & neurosurgery - Abstract
Background Models including an interaction term and performing a joint test of SNP and/or interaction effect are often used to discover Gene-Environment (GxE) interactions. When the environmental exposure is a binary variable, analyses from exposure-stratified models which consist of estimating genetic effect in unexposed and exposed individuals separately can be of interest. In large-scale consortia focusing on GxE interactions in which only the joint test has been performed, it may be challenging to get summary statistics from both exposure-stratified and marginal (i.e not accounting for interaction) models. Results In this work, we developed a simple framework to estimate summary statistics in each stratum of a binary exposure and in the marginal model using summary statistics from the “joint” model. We performed simulation studies to assess our estimators’ accuracy and examined potential sources of bias, such as correlation between genotype and exposure and differing phenotypic variances within exposure strata. Results from these simulations highlight the high theoretical accuracy of our estimators and yield insights into the impact of potential sources of bias. We then applied our methods to real data and demonstrate our estimators’ retained accuracy after filtering SNPs by sample size to mitigate potential bias. Conclusions These analyses demonstrated the accuracy of our method in estimating both stratified and marginal summary statistics from a joint model of gene-environment interaction. In addition to facilitating the interpretation of GxE screenings, this work could be used to guide further functional analyses. We provide a user-friendly Python script to apply this strategy to real datasets. The Python script and documentation are available at https://gitlab.pasteur.fr/statistical-genetics/j2s.
- Published
- 2019
- Full Text
- View/download PDF