Candidate Gene–Environment Interaction Research: Reflections and Recommendations

Danielle M. Dick; Arpana Agrawal; Matthew C. Keller; Amy Adkins; Fazil Aliev; Scott Monroe; John K. Hewitt; Kenneth S. Kendler; Kenneth J. Sher

doi:10.1177/1745691614556682

Candidate Gene–Environment Interaction Research: Reflections and Recommendations

Danielle M. Dick, Arpana Agrawal, Matthew C. Keller, Amy Adkins, Fazil Aliev, Scott Monroe, John K. Hewitt, Kenneth S. Kendler, Kenneth J. Sher

Research output: Contribution to journal › Article › peer-review

247 Scopus citations

Abstract

Studying how genetic predispositions come together with environmental factors to contribute to complex behavioral outcomes has great potential for advancing the understanding of the development of psychopathology. It represents a clear theoretical advance over studying these factors in isolation. However, research at the intersection of multiple fields creates many challenges. We review several reasons why the rapidly expanding candidate gene–environment interaction (cG×E) literature should be considered with a degree of caution. We discuss lessons learned about candidate gene main effects from the evolving genetics literature and how these inform the study of cG×E. We review the importance of the measurement of the gene and environment of interest in cG×E studies. We discuss statistical concerns with modeling cG×E that are frequently overlooked. Furthermore, we review other challenges that have likely contributed to the cG×E literature being difficult to interpret, including low power and publication bias. Many of these issues are similar to other concerns about research integrity (e.g., high false-positive rates) that have received increasing attention in the social sciences. We provide recommendations for rigorous research practices for cG×E studies that we believe will advance its potential to contribute more robustly to the understanding of complex behavioral phenotypes.

Original language	English
Pages (from-to)	37-59
Number of pages	23
Journal	Perspectives on Psychological Science
Volume	10
Issue number	1
DOIs	https://doi.org/10.1177/1745691614556682
State	Published - Jan 21 2015

Keywords

G×E
candidate genes
genetics
gene–environment interaction

Access to Document

10.1177/1745691614556682

Cite this

@article{1e4758c259ac442f8776679d22fd9eeb,

title = "Candidate Gene–Environment Interaction Research: Reflections and Recommendations",

abstract = "Studying how genetic predispositions come together with environmental factors to contribute to complex behavioral outcomes has great potential for advancing the understanding of the development of psychopathology. It represents a clear theoretical advance over studying these factors in isolation. However, research at the intersection of multiple fields creates many challenges. We review several reasons why the rapidly expanding candidate gene–environment interaction (cG×E) literature should be considered with a degree of caution. We discuss lessons learned about candidate gene main effects from the evolving genetics literature and how these inform the study of cG×E. We review the importance of the measurement of the gene and environment of interest in cG×E studies. We discuss statistical concerns with modeling cG×E that are frequently overlooked. Furthermore, we review other challenges that have likely contributed to the cG×E literature being difficult to interpret, including low power and publication bias. Many of these issues are similar to other concerns about research integrity (e.g., high false-positive rates) that have received increasing attention in the social sciences. We provide recommendations for rigorous research practices for cG×E studies that we believe will advance its potential to contribute more robustly to the understanding of complex behavioral phenotypes.",

keywords = "G×E, candidate genes, genetics, gene–environment interaction",

author = "Dick, {Danielle M.} and Arpana Agrawal and Keller, {Matthew C.} and Amy Adkins and Fazil Aliev and Scott Monroe and Hewitt, {John K.} and Kendler, {Kenneth S.} and Sher, {Kenneth J.}",

note = "Funding Information: Studying how genetic predispositions and environmental circumstances come together to contribute to complex behavioral outcomes has great potential for advancing the understanding of the development of psychopathology. It represents a clear theoretical advance over studying these factors in isolation. However, research at the intersection of multiple fields creates many challenges. Studying cG×E requires appropriate understanding of genetic mechanisms, appropriate measurement of the environment, as well as a conceptual framework for integrating the two with respect to a specific outcome of interest and, critically, of the statistical principles that underpin cG×E studies. It is not likely, or expected, that every investigator conducting cG×E research will have the requisite expertise in all of these areas; accordingly, we encourage cG×E studies that are collaborative efforts involving individuals with common interests but diverse expertise. The National Institutes of Health has initiatives aimed at addressing challenges associated with genotypic research, many of which are also relevant to the study of cG×E. For example, in recognition of the fact that gene identification requires large numbers, but one of the challenges that is often encountered as researchers attempt to pool their samples is the use of different measures across studies, the PhenX initiative was launched ( www.phenX.org ; Hamilton et al., 2011 ). PhenX brought together panels of experts across a variety of research areas to come up with recommended consensus measures (including both outcome and environmental measures) for inclusion in genetics studies to encourage the use of common measures to facilitate cross-study comparisons and analyses. There are limitations to this approach: Brief, low-burden measures were preferentially selected to encourage more widespread uptake, which may result in less precise or comprehensive assessments in the case of some constructs. However, it represents a step toward facilitating collaborative efforts in genetics research. The use of standardized measures across studies could also help advance the cG×E field, with greater emphasis placed on replication and combined analyses across research groups to enhance sample size and corresponding power. The Office of Behavioral and Social Sciences Research at the National Institutes of Health also has resources on its website ( http://obssr.od.nih.gov/ ) to help social scientists with the incorporation of genetic information into their studies. Many of these issues are not specific to the study of psychopathology. In a recent article on G×Es in cancer epidemiology, coming out of a National Cancer Institute think tank, Hutter, Mechanic, Chatterjee, Kraft, and Gillanders (2013) described similar challenges. As investigators who have explored cG×E hypotheses ourselves, with some of our own work not meeting the standards delineated in this review, we were compelled to ask “how can we do better?” We hope we have outlined some such strategies. Through greater awareness of the challenges in conducting cG×E research, resources available to aid in conducting high-quality cG×E studies, and proactive efforts to move cG×E studies in this direction, it is our hope that this growing area of research will eventually reach its potential to deeply inform to the understanding of complex behavioral outcomes. This article developed out of discussions from the workshop “Challenges and Opportunities in G×E Research: Creating Consensus Recommendations for Guidelines on G×E Research,” sponsored by the National Institute on Alcohol Abuse and Alcoholism (NIAAA) in January 2013. In addition to the authors who contributed to the writing of this article, other workshop participants included Lindon Eaves (Virginia Commonwealth University), Andrew Heath (Washington University in St. Louis), and Eric Turkheimer (University of Virginia). We are deeply grateful for the leadership of Marcia S. Scott (chair of the NIAAA Gene and Environment Team), Dionne C. Godette, and Mariela C. Shirley—who are affiliated with the Division of Epidemiology and Prevention Research, NIAAA (Mariela C. Shirley is now at the Office of Research on Women{\textquoteright}s Health), National Institutes of Health—in organizing support for the workshop and for their helpful comments on drafts of this article. Declaration of Conflicting Interests The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article. Funding Support for the authors involved in writing this article includes the following: K02AA018755, R01DA070312, and U10AA008401 to Danielle M. Dick; R01AA018333 to Danielle M. Dick and Kenneth S. Kendler; K01MH085812 and R01MH100141 to Matthew C. Keller; K02DA32573, R21AA21235, and R01DA23668 to Arpana Agrawal; DA011015 to John K. Hewitt; and K05 AA017242 to Kenneth J. Sher. We are also grateful to Erica Spotts, Health Scientist Administrator at the Office of Behavioral and Social Science Research (OBSSR), for her assistance in securing funding (R01 AA018333-04S1 to Danielle M. Dick) to create a website to facilitate understanding of the concepts related to conducting G×EcG×E research, as delineated in this article. This information will be hosted through the OBSSR website (anticipated 2015 release). Supplemental Material Additional supporting information may be found at http://pps.sagepub.com/content/by/supplemental-data 1. Words in bold font in the text are defined in the Supplemental Material. 2. This is often called LD-tagging and indicates that you can use information about the LD or correlation structure across variants within and across genes to know how many genetic variants you need to genotype to capture most of the variable locations in the gene that could be associated with outcome (see Text Box 1 in the Supplemental Material for software). 3. Mendelian randomization is a related and novel conceptualization of genotype representing environmental exposure propensity and might explain cG×E in the presence of rGE. We refer the reader to Smith (2010) to learn more. 4. Passive rGE is more common during childhood and adolescence and refers to individuals being differentially exposed to an environment without their own initiative, most likely because aspects of their environment are provided by their parents with whom they also share genetic variance. For instance, antisocial parents may pass on a genetic liability to antisociality and expose their children to an abusive environment. Evocative rGE refers to environments that an individual elicits/evokes from others on the basis of his or her genetic predisposition. For instance, an antisocial adolescent may evoke harsher parenting. Finally, as an individual matures into adulthood, the importance of active rGE increases. Here, an individual actively selects his or her environment, or niche, on the basis of characteristics of his or her genetic predisposition. For example, antisocial youths may select into high-risk neighborhoods or affiliate with deviant peers. 5. The ADH1B polymorphism (rs1229984) has been implicated in Asian populations to afford protection against the development of alcoholism, putatively via flushing (facial reddening due to accelerated conversion of ethanol to acetaldehyde); rs1229984 was not originally detected in GWAS on European American populations. The minor allele frequency of this variant in European Americans is low (<5%), and commercial GWAS arrays rely on common variation. Accordingly, the SNP was neither genotyped on these arrays nor could it be reliably imputed. However, genotyping this polymorphism resulted in genome-wide significant association signals ( Bierut et al., 2012 ), and with the advent of more recent arrays that target this variant better, there is now evidence in GWAS of an association between rs1229984 and alcoholism at p = 1.2 × 10 −31 ( Gelernter et al., 2014 ). In this instance, GWAS did not initially guide identification of a logical and validated candidate gene, highlighting that all techniques have their strengths and weaknesses. Publisher Copyright: {\textcopyright} The Author(s) 2014.",

year = "2015",

month = jan,

day = "21",

doi = "10.1177/1745691614556682",

language = "English",

volume = "10",

pages = "37--59",

journal = "Perspectives on Psychological Science",

issn = "1745-6916",

number = "1",

}

TY - JOUR

T1 - Candidate Gene–Environment Interaction Research

T2 - Reflections and Recommendations

AU - Dick, Danielle M.

AU - Agrawal, Arpana

AU - Keller, Matthew C.

AU - Adkins, Amy

AU - Aliev, Fazil

AU - Monroe, Scott

AU - Hewitt, John K.

AU - Kendler, Kenneth S.

AU - Sher, Kenneth J.

N1 - Funding Information: Studying how genetic predispositions and environmental circumstances come together to contribute to complex behavioral outcomes has great potential for advancing the understanding of the development of psychopathology. It represents a clear theoretical advance over studying these factors in isolation. However, research at the intersection of multiple fields creates many challenges. Studying cG×E requires appropriate understanding of genetic mechanisms, appropriate measurement of the environment, as well as a conceptual framework for integrating the two with respect to a specific outcome of interest and, critically, of the statistical principles that underpin cG×E studies. It is not likely, or expected, that every investigator conducting cG×E research will have the requisite expertise in all of these areas; accordingly, we encourage cG×E studies that are collaborative efforts involving individuals with common interests but diverse expertise. The National Institutes of Health has initiatives aimed at addressing challenges associated with genotypic research, many of which are also relevant to the study of cG×E. For example, in recognition of the fact that gene identification requires large numbers, but one of the challenges that is often encountered as researchers attempt to pool their samples is the use of different measures across studies, the PhenX initiative was launched ( www.phenX.org ; Hamilton et al., 2011 ). PhenX brought together panels of experts across a variety of research areas to come up with recommended consensus measures (including both outcome and environmental measures) for inclusion in genetics studies to encourage the use of common measures to facilitate cross-study comparisons and analyses. There are limitations to this approach: Brief, low-burden measures were preferentially selected to encourage more widespread uptake, which may result in less precise or comprehensive assessments in the case of some constructs. However, it represents a step toward facilitating collaborative efforts in genetics research. The use of standardized measures across studies could also help advance the cG×E field, with greater emphasis placed on replication and combined analyses across research groups to enhance sample size and corresponding power. The Office of Behavioral and Social Sciences Research at the National Institutes of Health also has resources on its website ( http://obssr.od.nih.gov/ ) to help social scientists with the incorporation of genetic information into their studies. Many of these issues are not specific to the study of psychopathology. In a recent article on G×Es in cancer epidemiology, coming out of a National Cancer Institute think tank, Hutter, Mechanic, Chatterjee, Kraft, and Gillanders (2013) described similar challenges. As investigators who have explored cG×E hypotheses ourselves, with some of our own work not meeting the standards delineated in this review, we were compelled to ask “how can we do better?” We hope we have outlined some such strategies. Through greater awareness of the challenges in conducting cG×E research, resources available to aid in conducting high-quality cG×E studies, and proactive efforts to move cG×E studies in this direction, it is our hope that this growing area of research will eventually reach its potential to deeply inform to the understanding of complex behavioral outcomes. This article developed out of discussions from the workshop “Challenges and Opportunities in G×E Research: Creating Consensus Recommendations for Guidelines on G×E Research,” sponsored by the National Institute on Alcohol Abuse and Alcoholism (NIAAA) in January 2013. In addition to the authors who contributed to the writing of this article, other workshop participants included Lindon Eaves (Virginia Commonwealth University), Andrew Heath (Washington University in St. Louis), and Eric Turkheimer (University of Virginia). We are deeply grateful for the leadership of Marcia S. Scott (chair of the NIAAA Gene and Environment Team), Dionne C. Godette, and Mariela C. Shirley—who are affiliated with the Division of Epidemiology and Prevention Research, NIAAA (Mariela C. Shirley is now at the Office of Research on Women’s Health), National Institutes of Health—in organizing support for the workshop and for their helpful comments on drafts of this article. Declaration of Conflicting Interests The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article. Funding Support for the authors involved in writing this article includes the following: K02AA018755, R01DA070312, and U10AA008401 to Danielle M. Dick; R01AA018333 to Danielle M. Dick and Kenneth S. Kendler; K01MH085812 and R01MH100141 to Matthew C. Keller; K02DA32573, R21AA21235, and R01DA23668 to Arpana Agrawal; DA011015 to John K. Hewitt; and K05 AA017242 to Kenneth J. Sher. We are also grateful to Erica Spotts, Health Scientist Administrator at the Office of Behavioral and Social Science Research (OBSSR), for her assistance in securing funding (R01 AA018333-04S1 to Danielle M. Dick) to create a website to facilitate understanding of the concepts related to conducting G×EcG×E research, as delineated in this article. This information will be hosted through the OBSSR website (anticipated 2015 release). Supplemental Material Additional supporting information may be found at http://pps.sagepub.com/content/by/supplemental-data 1. Words in bold font in the text are defined in the Supplemental Material. 2. This is often called LD-tagging and indicates that you can use information about the LD or correlation structure across variants within and across genes to know how many genetic variants you need to genotype to capture most of the variable locations in the gene that could be associated with outcome (see Text Box 1 in the Supplemental Material for software). 3. Mendelian randomization is a related and novel conceptualization of genotype representing environmental exposure propensity and might explain cG×E in the presence of rGE. We refer the reader to Smith (2010) to learn more. 4. Passive rGE is more common during childhood and adolescence and refers to individuals being differentially exposed to an environment without their own initiative, most likely because aspects of their environment are provided by their parents with whom they also share genetic variance. For instance, antisocial parents may pass on a genetic liability to antisociality and expose their children to an abusive environment. Evocative rGE refers to environments that an individual elicits/evokes from others on the basis of his or her genetic predisposition. For instance, an antisocial adolescent may evoke harsher parenting. Finally, as an individual matures into adulthood, the importance of active rGE increases. Here, an individual actively selects his or her environment, or niche, on the basis of characteristics of his or her genetic predisposition. For example, antisocial youths may select into high-risk neighborhoods or affiliate with deviant peers. 5. The ADH1B polymorphism (rs1229984) has been implicated in Asian populations to afford protection against the development of alcoholism, putatively via flushing (facial reddening due to accelerated conversion of ethanol to acetaldehyde); rs1229984 was not originally detected in GWAS on European American populations. The minor allele frequency of this variant in European Americans is low (<5%), and commercial GWAS arrays rely on common variation. Accordingly, the SNP was neither genotyped on these arrays nor could it be reliably imputed. However, genotyping this polymorphism resulted in genome-wide significant association signals ( Bierut et al., 2012 ), and with the advent of more recent arrays that target this variant better, there is now evidence in GWAS of an association between rs1229984 and alcoholism at p = 1.2 × 10 −31 ( Gelernter et al., 2014 ). In this instance, GWAS did not initially guide identification of a logical and validated candidate gene, highlighting that all techniques have their strengths and weaknesses. Publisher Copyright: © The Author(s) 2014.

PY - 2015/1/21

Y1 - 2015/1/21

N2 - Studying how genetic predispositions come together with environmental factors to contribute to complex behavioral outcomes has great potential for advancing the understanding of the development of psychopathology. It represents a clear theoretical advance over studying these factors in isolation. However, research at the intersection of multiple fields creates many challenges. We review several reasons why the rapidly expanding candidate gene–environment interaction (cG×E) literature should be considered with a degree of caution. We discuss lessons learned about candidate gene main effects from the evolving genetics literature and how these inform the study of cG×E. We review the importance of the measurement of the gene and environment of interest in cG×E studies. We discuss statistical concerns with modeling cG×E that are frequently overlooked. Furthermore, we review other challenges that have likely contributed to the cG×E literature being difficult to interpret, including low power and publication bias. Many of these issues are similar to other concerns about research integrity (e.g., high false-positive rates) that have received increasing attention in the social sciences. We provide recommendations for rigorous research practices for cG×E studies that we believe will advance its potential to contribute more robustly to the understanding of complex behavioral phenotypes.

AB - Studying how genetic predispositions come together with environmental factors to contribute to complex behavioral outcomes has great potential for advancing the understanding of the development of psychopathology. It represents a clear theoretical advance over studying these factors in isolation. However, research at the intersection of multiple fields creates many challenges. We review several reasons why the rapidly expanding candidate gene–environment interaction (cG×E) literature should be considered with a degree of caution. We discuss lessons learned about candidate gene main effects from the evolving genetics literature and how these inform the study of cG×E. We review the importance of the measurement of the gene and environment of interest in cG×E studies. We discuss statistical concerns with modeling cG×E that are frequently overlooked. Furthermore, we review other challenges that have likely contributed to the cG×E literature being difficult to interpret, including low power and publication bias. Many of these issues are similar to other concerns about research integrity (e.g., high false-positive rates) that have received increasing attention in the social sciences. We provide recommendations for rigorous research practices for cG×E studies that we believe will advance its potential to contribute more robustly to the understanding of complex behavioral phenotypes.

KW - G×E

KW - candidate genes

KW - genetics

KW - gene–environment interaction

UR - http://www.scopus.com/inward/record.url?scp=84921528220&partnerID=8YFLogxK

U2 - 10.1177/1745691614556682

DO - 10.1177/1745691614556682

M3 - Article

C2 - 25620996

AN - SCOPUS:84921528220

SN - 1745-6916

VL - 10

SP - 37

EP - 59

JO - Perspectives on Psychological Science

JF - Perspectives on Psychological Science

IS - 1

ER -

Candidate Gene–Environment Interaction Research: Reflections and Recommendations

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this