Why we need families in genomic research on developmental psychopathology

Abstract

Background: Fundamental questions about the roles of genes, environments, and their interplay in developmental psychopathology have traditionally been the domain of twin and family studies. More recently, the rapidly growing availability of large genomic datasets, composed of unrelated individuals, has generated novel insights. However, there are major stumbling blocks. Only a small fraction of the total genetic influence on childhood psychopathology estimated from family data is captured with measured DNA. Moreover, genetic influence identified using DNA is often confounded with indirect genetic effects of relatives, population stratification and assortative mating.

Methods: The goal of this paper is to review how combining DNA-based genomic research with family-based quantitative genetics helps to address key issues in genomics and push knowledge further.

Results: We focus on three approaches to obtaining more accurate and novel genomic findings on the developmental aetiology of psychopathology: (a) using knowledge from twin and family studies, (b) triangulating with twin and family studies, and (c) integrating data and methods with twin and family studies.

Conclusion: We support the movement towards family-based genomic research, and show that developmental psychologists are particularly well-placed to contribute hypotheses, analysis tools, and data.

Keywords: genomics; psychopathology; twin and family studies.

FIGURE 1

Estimating heritability. Heritability is the proportion of phenotypic variance among individuals that can be attributed to genetic differences in a certain population. Figure 1 sketches the heritability of an imagined phenotype in an imagined population (left, composed of direct effects of common and rare genetic variants), and how accurately population‐based SNP heritability estimates (middle) and classical MZ‐DZ twin heritability estimates (right) are likely to capture this under certain circumstances. SNP heritability estimates include only genetic effects tagged by SNPs included in the analysis, whilst twin heritability estimates capture effects of all DNA differences in the sample. Various factors can lead heritability estimates to be overestimated (red circles increasing the size of the estimate) or underestimated (white circles representing empty space reducing the size of the estimate). The impact of each biasing factor is population and phenotype‐dependent (the number of red vs. white circles would vary across phenotypes), and factors do not necessarily work in the same way for SNP‐ and twin‐based methods (some factors are red for SNP heritability but white for twin heritability e.g., assortative mating). SNP heritability estimates can be inflated by indirect genetic effects, assortative mating, and population stratification (red), but deflated if genetic effects covary negatively with indirect genetic effects (white). Twin heritability estimates can be inflated if the equal environments assumption does not hold (i.e., greater environmental sharing among MZ than DZ twins) and in the presence of non‐additive genetic effects and gene‐by‐shared environment interaction and deflated in the presence of assortative mating. Missing heritability (dashed grey arrow) is the gap between SNP‐and twin‐based heritability estimates. A key explanation for missing heritability is that SNP heritability estimates cannot capture rare genetic effects. Deflating and inflating factors make it difficult to assess missing heritability. For example, the missing heritability problem could be more severe than assumed if assortative mating is inflating SNP heritability estimates and deflating twin heritability estimates.

FIGURE 2

Genetic, environmental and confounding influences on childhood psychopathology. Note that G and E represent the genetic and environmental influences on a child's trait. A direct genetic effect is the effect of the child's genetic variation on their own trait, through the pathway: Mother/Father –> G –> Child's trait. As well as transmitting genes, parents provide part of the child's environment. These genetic and environmental influences are not independent, because the environmental effect on the child trait may be partially influenced by parental genotype. Indeed, an indirect parental genetic effect is an environmentally‐mediated effect of the parental genome on the offspring phenotype, such as when parental genetic risk for depression impacts on child depression via parental emotional symptoms (Cheesman, Eilertsen, et al., 2020). Indirect parental genetic effects are included in the diagram through the following pathways: Mother/Father –> E –> Child's trait. Note that whilst PGS for non‐transmitted alleles can be used to capture indirect parental genetic effects (Kong et al., 2018), they are not the same thing, since the former can also arise from transmitted parental alleles. The path from G to E represents evocative and active gene‐environment correlation, and the presence of paths from Mother/Father to both G and E reflect passive gene‐environment correlation. Assortative mating occurs when there is greater similarity between partners than expected by chance and is reflected by the double headed arrows running between mother and father. When mating is influenced by heritable characteristics, this results in increased trait‐specific genetic and phenotypic variance in the child generation. Population stratification can be described as confounding due to ancestry differences in the population. Genetic differences between subpopulations (with different allele frequencies) can become correlated with phenotypic differences (in parents and/or children) even if they do not have a causal effect on the trait.

FIGURE 3

Leveraging family data in genomics in three ways. The present article puts forward three ways in which family data can be used to advance knowledge on the genomics of childhood psychopathology: using knowledge, triangulating, and integrating. For example, knowledge from twin research on gene‐environment correlation can help us to interpret population‐based genomic associations with psychopathology and generate testable hypotheses about mediating mechanisms. Triangulating twin and genomic methods with different assumptions in the same dataset can, for example, allow us to quantify the role of parental behaviour relative to other environmental influences on psychopathology. Integrating family and genomic data can help to recover missing heritability by capturing rare genetic effects that are missed in population‐based studies of unrelated people.