The target article is to be commended for highlighting exciting new developments in molecular genetics, but its dismissal of the better-established methods in behavioral genetics research misses the big picture, the goals of these methods. The biometric analysis of behavioral phenotypes (“heritability studies”) was never meant to specify a theory of human behavior. Rather, biometry provides a useful, and approximate, characterization of the nature of familial resemblance that implicates the importance of heritable factors and provides a foundation for targeted examination of those factors. Similarly, the investigation of genotype–phenotype association is not intended to specify the exact mechanisms by which the genotype is translated into the phenotype, but rather to examine the phenotypic consequences of variation in DNA. Research on mechanisms intermediary between genotype and phenotype builds on rather than competes with genetic association studies. All models of family resemblance and genotype–phenotype relations are imperfect, but that does not mean they are not useful.
Charney claims, without citing any evidence, that “heritability and gene association studies depend” on “six basic assumptions” (HS1–HS3, for heritability studies, and GAS1–GAS3 for genetic association studies).
Assumption HS1 asserts that heritability estimation assumes MZ (monozygotic) twins are genetically identical. Charney aims to refute this assumption by referring to studies that suggest MZ twins ought to differ in their DNA in somatic cells and in mitochondrial DNA. Most of the MZ twin differences he cites have nothing to do with heritability because they are acquired and are not differences in inherited DNA. We have known about genetic differences between MZ twins for decades (Zwijnenburg et al. Reference Zwijnenburg, Meijers-Heijboer and Boomsma2010), but we also know that MZ twins rarely differ in measured genotypes. CNVs (copy number variations) and methylation patterns are more promising fronts that are being vigorously explored as a basis for MZ twin discordance (Bell & Spector Reference Bell and Spector2011).
Charney's definition and discussion of HS2 and HS3 reveal his misunderstanding of biometric studies. Heritability indicates the percentage of variance in a phenotype that is due to inheritance, not due to variations in DNA. There is an important difference because, as Charney points out, many new variations in DNA may arise during development and certain epigenetic changes are heritable. The first model for heritability estimation was proposed by Sewall Wright (Reference Wright1920), years before the importance of DNA to genetics was understood. Biometrical models have become progressively more sophisticated over the years, and as new kinds of genetic and environmental phenomena are discovered, their effects can be incorporated into the models. Models include much more than G, E and G × E. We also model age and sex effects on heritability (Haworth et al. Reference Haworth, Wright, Luciano, Martin, de Geus, van Beijsterveldt, Bartels, Posthuma, Boomsma, Davis, Kovas, Corley, DeFries, Hewitt, Olson, Rhea, Wadsworth, Iacono, McGue, Thompson, Hart, Petrill, Lubinski and Plomin2010), effects of assortative mating (Vinkhuyzen et al. Reference Vinkhuyzen, van der Sluis, Maes and Posthuma2012), the prenatal environment (Devlin et al. Reference Devlin, Daniels and Roeder1997), sex-specific effects, social interactions such as sibling competition and sibling cooperation, measured genotypes, and measured environment effects (Neale & Cardon Reference Neale and Cardon1992). Factors that cannot be modeled well today because we lack good data, such as cytosine methylation or retrotransposons, may be incorporated someday and the effects of these factors on model parameters can already be investigated through simulation.
Like HS1, GAS1 (persons having identical DNA in all cells) need only be accurate in a limited sense for genetic association studies to be valid, and considerable evidence supports this validity. We typically extract DNA from white blood cells or from cells in saliva. DNA in those cells is not necessarily the same as DNA in other cells. However, several types of evidence suggest that DNA extracted from different cell types is nearly identical. MZ twins are nearly as alike in their SNP (single nucleotide polymorphism) genotypes as are two DNA samples from the same person; we see the expected patterns of haplotype sharing for pairs of relatives (e.g., siblings or parent–offspring pairs); and there is very little evidence of Mendelian inconsistency in GWAS (genome-wide association studies) of parents and their offspring. In all of these cases, the differences we observe occur at the rate we would expect due to genotyping error, indicating that the rare changes that do occur are inconsequential for an association study.
Neither GAS2 nor GAS3 are assumed in genetic association research. In typical genotype–phenotype association research (including GWAS), one tests the association of a measured genotype with a phenotype. This method has very few assumptions, only an association test followed by the challenge of interpreting the result. If an association is found, this means that genotype predicts phenotype regardless of epigenetic moderation of gene expression or any of the intermediary processes that Charney describes. Genetic association studies should not be abandoned, even if they are made more powerful through the addition of analyses of CNVs, gene expression, and so forth.
Charney's greatest mistake is his misunderstanding of the role of quantitative genetic models. One of the most salient features of human behavior is that it is familial. Indeed, despite Charney's focus on differences, MZ twins are strikingly similar: correlated approximately .80 for IQ (Bouchard & McGue Reference Bouchard and McGue1981), .50 on diverse personality characteristics (Bouchard & Loehlin Reference Bouchard and Loehlin2001), and for a trait of particular interest to political scientists like Charney, .50 in their political ideologies (Alford et al. Reference Alford, Funk and Hibbing2005). It is hard to imagine that the investigation of how MZ twins come to be so similar would provide no insights into the nature of the development of these fundamental human traits. Quantitative genetics is a powerful tool for initiating a program of research aimed ultimately at explicating the mechanisms leading to familial resemblance: Behavior is heritable, implying inherited differences in DNA sequence contribute in some way to behavioral differences. DNA sequence markers are associated (albeit weakly) with some behavioral phenotypes, implying that variants in the region of these markers are likely causal. It is not that the phenomena that draw Charney's interest – retrotransposons, mosaicism, and structural variation – are unimportant. They clearly are important. Rather, what these diverse phenomena do not provide is a coherent alternative paradigm for what remains one of the most important observations about human behavior, namely, that it is transmitted within families in a predictable manner.
The target article is to be commended for highlighting exciting new developments in molecular genetics, but its dismissal of the better-established methods in behavioral genetics research misses the big picture, the goals of these methods. The biometric analysis of behavioral phenotypes (“heritability studies”) was never meant to specify a theory of human behavior. Rather, biometry provides a useful, and approximate, characterization of the nature of familial resemblance that implicates the importance of heritable factors and provides a foundation for targeted examination of those factors. Similarly, the investigation of genotype–phenotype association is not intended to specify the exact mechanisms by which the genotype is translated into the phenotype, but rather to examine the phenotypic consequences of variation in DNA. Research on mechanisms intermediary between genotype and phenotype builds on rather than competes with genetic association studies. All models of family resemblance and genotype–phenotype relations are imperfect, but that does not mean they are not useful.
Charney claims, without citing any evidence, that “heritability and gene association studies depend” on “six basic assumptions” (HS1–HS3, for heritability studies, and GAS1–GAS3 for genetic association studies).
Assumption HS1 asserts that heritability estimation assumes MZ (monozygotic) twins are genetically identical. Charney aims to refute this assumption by referring to studies that suggest MZ twins ought to differ in their DNA in somatic cells and in mitochondrial DNA. Most of the MZ twin differences he cites have nothing to do with heritability because they are acquired and are not differences in inherited DNA. We have known about genetic differences between MZ twins for decades (Zwijnenburg et al. Reference Zwijnenburg, Meijers-Heijboer and Boomsma2010), but we also know that MZ twins rarely differ in measured genotypes. CNVs (copy number variations) and methylation patterns are more promising fronts that are being vigorously explored as a basis for MZ twin discordance (Bell & Spector Reference Bell and Spector2011).
Charney's definition and discussion of HS2 and HS3 reveal his misunderstanding of biometric studies. Heritability indicates the percentage of variance in a phenotype that is due to inheritance, not due to variations in DNA. There is an important difference because, as Charney points out, many new variations in DNA may arise during development and certain epigenetic changes are heritable. The first model for heritability estimation was proposed by Sewall Wright (Reference Wright1920), years before the importance of DNA to genetics was understood. Biometrical models have become progressively more sophisticated over the years, and as new kinds of genetic and environmental phenomena are discovered, their effects can be incorporated into the models. Models include much more than G, E and G × E. We also model age and sex effects on heritability (Haworth et al. Reference Haworth, Wright, Luciano, Martin, de Geus, van Beijsterveldt, Bartels, Posthuma, Boomsma, Davis, Kovas, Corley, DeFries, Hewitt, Olson, Rhea, Wadsworth, Iacono, McGue, Thompson, Hart, Petrill, Lubinski and Plomin2010), effects of assortative mating (Vinkhuyzen et al. Reference Vinkhuyzen, van der Sluis, Maes and Posthuma2012), the prenatal environment (Devlin et al. Reference Devlin, Daniels and Roeder1997), sex-specific effects, social interactions such as sibling competition and sibling cooperation, measured genotypes, and measured environment effects (Neale & Cardon Reference Neale and Cardon1992). Factors that cannot be modeled well today because we lack good data, such as cytosine methylation or retrotransposons, may be incorporated someday and the effects of these factors on model parameters can already be investigated through simulation.
Like HS1, GAS1 (persons having identical DNA in all cells) need only be accurate in a limited sense for genetic association studies to be valid, and considerable evidence supports this validity. We typically extract DNA from white blood cells or from cells in saliva. DNA in those cells is not necessarily the same as DNA in other cells. However, several types of evidence suggest that DNA extracted from different cell types is nearly identical. MZ twins are nearly as alike in their SNP (single nucleotide polymorphism) genotypes as are two DNA samples from the same person; we see the expected patterns of haplotype sharing for pairs of relatives (e.g., siblings or parent–offspring pairs); and there is very little evidence of Mendelian inconsistency in GWAS (genome-wide association studies) of parents and their offspring. In all of these cases, the differences we observe occur at the rate we would expect due to genotyping error, indicating that the rare changes that do occur are inconsequential for an association study.
Neither GAS2 nor GAS3 are assumed in genetic association research. In typical genotype–phenotype association research (including GWAS), one tests the association of a measured genotype with a phenotype. This method has very few assumptions, only an association test followed by the challenge of interpreting the result. If an association is found, this means that genotype predicts phenotype regardless of epigenetic moderation of gene expression or any of the intermediary processes that Charney describes. Genetic association studies should not be abandoned, even if they are made more powerful through the addition of analyses of CNVs, gene expression, and so forth.
Charney's greatest mistake is his misunderstanding of the role of quantitative genetic models. One of the most salient features of human behavior is that it is familial. Indeed, despite Charney's focus on differences, MZ twins are strikingly similar: correlated approximately .80 for IQ (Bouchard & McGue Reference Bouchard and McGue1981), .50 on diverse personality characteristics (Bouchard & Loehlin Reference Bouchard and Loehlin2001), and for a trait of particular interest to political scientists like Charney, .50 in their political ideologies (Alford et al. Reference Alford, Funk and Hibbing2005). It is hard to imagine that the investigation of how MZ twins come to be so similar would provide no insights into the nature of the development of these fundamental human traits. Quantitative genetics is a powerful tool for initiating a program of research aimed ultimately at explicating the mechanisms leading to familial resemblance: Behavior is heritable, implying inherited differences in DNA sequence contribute in some way to behavioral differences. DNA sequence markers are associated (albeit weakly) with some behavioral phenotypes, implying that variants in the region of these markers are likely causal. It is not that the phenomena that draw Charney's interest – retrotransposons, mosaicism, and structural variation – are unimportant. They clearly are important. Rather, what these diverse phenomena do not provide is a coherent alternative paradigm for what remains one of the most important observations about human behavior, namely, that it is transmitted within families in a predictable manner.