Large-scale genomics unveils the genetic architecture of psychiatric disorders

Abstract

Family study results are consistent with genetic effects making substantial contributions to risk of psychiatric disorders such as schizophrenia, yet robust identification of specific genetic variants that explain variation in population risk had been disappointing until the advent of technologies that assay the entire genome in large samples. We highlight recent progress that has led to a better understanding of the number of risk variants in the population and the interaction of allele frequency and effect size. The emerging genetic architecture implies a large number of contributing loci (that is, a high genome-wide mutational target) and suggests that genetic risk of psychiatric disorders involves the combined effects of many common variants of small effect, as well as rare and de novo variants of large effect. The capture of a substantial proportion of genetic risk facilitates new study designs to investigate the combined effects of genes and the environment.

Figure 1

The trajectory of GWAS discoveries for schizophrenia and other psychiatric disorders in comparison to Crohn’s disease and inflammatory bowel disease. The number of independent genome-wide significant (P < 5 × 10–8) loci are plotted as a function of the numbers of cases in the largest meta/mega-analysis. The inset shows detail for studies with fewer than 6,000 cases. CD-IBD, Crohn’s disease and inflammatory bowel disease (including CD and ulcerative colitis); Cross-disorder: broad psychiatric phenotype spanning ASD, ADHD, BPD, MDD, SCZ; SCZ-BPD, SCZ and BPD.

Figure 2

Not all GWASs are created equal under a polygenic architecture. For the same sample size, less common diseases have more power. Statistical power for detection of gene variants associated with disease for different disorders given the same sample size (type-1 error of 10–8, sample size of 10,000 cases and 10,000 controls). Different combinations of allele frequency and effect size can generate the same variance in liability. Examples of disorders with the prevalence shown are motor neuron disease (MND), SCZ and MDD.

Figure 3

Genetic discoveries for SCZ, irrespective of risk allele frequency, variant type (SNP or CNV) or discovery method (GWAS or CNV analysis), explain approximately the same proportion of the genetic variance. The red line defines a constant r2 of 0.04%, assuming a 1% prevalence of schizophrenia. Note that the population prevalence of some CNVs is set to 10–4 for convenience; some have not been observed in healthy controls and so the true allele frequency (and variance explained) will be lower.

Figure 4

Paternal age at child’s conception is associated with the burden of de novo mutations in the child’s genome (Poisson regression, P < 2 × 10–16, linear slope = 1.75 mutations per year). Data are from refs. (N = 78) and (N = 10).

Figure 5

Quantifying the genetic relationship between independent data sets through the SNP correlation^,,,. AA, African American; CHN, Chinese; EUR, European including European American.