Article by Frank Dudbridge, Pathways node of NCRM, London School of Hygiene and Tropical Medicine. This article also appears in the Spring 2012 issue of the MethodsNews newsletter.
Â
At first glance, genetics and social science lie at opposite ends of a spectrum. Our genes are fixed at conception and irrevocably determine our individual potentials for a multitude of physical, medical, and behavioural outcomes.
Social science, by contrast, is concerned with actions and results in a fluid world of human interaction, in which individual outcomes can be modified by judicious intervention. And yet the explosion of genetic data in the last decade has opened up many possibilities. Genetic data are available in many of the major cohorts that have been curated in social and medical research. What can genetics do for social scientists in the post-genomic era?
The two fields have long shared an uneasy common ground in heritability studies. These involve calculations expressing the correlations of a trait among relatives in terms of their stated relationships. With whole genome data we can now measure the exact genetic similarity between a pair of relatives. For example, siblings share half their genetic material on average, but a specific pair may by chance be up to 100% identical, and such siblings will be more similar for a heritable trait than the average. This leads to improved precision in heritability studies1. In epidemiological studies, genetic data allow us to build more complete models relating observations, and identify interactions between genes and social factors. For example, rural/urban environment is a possible effect modifier for the FTO gene in obesity2.
The fact that our genes are random given our parents, and the known causal direction from gene to outcome, has led to much interest in genetics for inferring causal relationships, using the method of instrumental variables first developed in econometrics. If we are interested in whether a certain trait causes an outcome, and we know that a gene influences that trait, we can substitute the gene for the trait in the analysis.
An association between the gene and the outcome then implies a causal relation between the trait and the outcome. For example, we may be interested in whether increased alcohol consumption causes an increased risk of heart disease. An association between alcohol and heart disease could be seen if, say, people who smoke are more likely to drink, but this does not imply that drinking causes heart disease in itself. However, certain genes are known to influence the level of alcohol consumption through their action on metabolism: carriers of particular genetic variants have more severe reactions to alcohol and tend to drink less as a result. An association between those genes and the risk of heart disease would imply that alcohol has a causal effect on disease, since it is unlikely that the genetic association could be explained by, say, smoking behaviour. This “Mendelian randomisation†approach mirrors the random allocation of patients in a clinical trial, and is a promising method for allowing causal conclusions to be drawn in social research3.
Through these new applications and common datasets, much greater collaboration between social scientists and geneticists is likely in the near future.
The Pathways node of the NCRM is running a series of courses to help social scientists become acquainted with the concepts and terminology of genetic and biomarker data. For details on upcoming courses please visit Pathways website.
References
1 Visscher PM, Hill WG, Wray NR (2008). Heritability in the genomics era - concepts and misconceptions. Nat Rev Genet 9: 255-266.
2 Taylor AE, Sandeep MN, Janipalli CS, Giambartolomei C, Evans DM, et al. (2011). Associations of FTO and MC4R Variants with Obesity Traits in Indians and the Role of Rural/Urban Environment as a Possible Effect Modifier. J Obes 2011: 307542.
3 Lawlor DA, Harbord RM, Sterne JA, Timpson N, Davey Smith G (2008). Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Stat Med 27: 1133-1163.
Â