Heritability estimates for psychiatric disorders such as bipolar (85%), unipolar disorder (59%), alcoholism (50–60%), attention deficit hyperactivity disorder (ADHD; 76%), and schizophrenia (85%) [Anon, 1983; McGuffin et al., 2003; Faraone et al., 2005; Husted et al., 2006; Stacey et al., 2009; Rommelse et al., 2010; Hallmayer et al., 2011] are notably higher than for other complex diseases such as cancer and several measures of heart disease, which typically have heritability estimates less than 60% (25–57%) [Austin et al., 1987; Sluijter et al., 1989; Page et al., 1997]. Traditional statistical approaches to gene identification for psychiatric disorders have used genetic markers to identify genetic variants that are associated with disease. This started with single markers, expanded to 100's of markers in linkage analysis, evolved to 100's of thousands of markers in genome-wide association, and has now become whole genome and exome sequencing. This approach has had some success for psychiatric disorders, most notably leading to the identification of APOE for Alzheimer's disease [Corder et al., 1995]. Both common [Stefansson et al., 2009; Jia et al., 2010; Ripke et al., 2011] and rare [Stefansson et al., 2008] genetic variants have also been identified for schizophrenia while primarily common genetic variants have been identified for bipolar disorder [Sklar et al., 2011]. However, meta-analyses for both major depressive disorder and ADHD have found no genetic variants that meet genome-wide significance using genome-wide association approaches [Neale et al., 2010; Ripke et al., 2013]. Even given the identified genetic variants for psychiatric disorders to date, these findings explain a shockingly small about of the total genetic variation [Manolio et al., 2009]. In addition, the understanding of the functional role that these variants have played in the pathogenesis of disease is most often not understood, making it difficult for any of these polymorphisms to have a clear therapeutic impact on the disorder. By itself, it is clear that the single marker approach is limited in the knowledge it can provide towards a complete understanding of the molecular mechanisms underlying psychiatric diseases for several reasons:
- Small genetic effect: To date the identified psychiatric genetic variants have most often only had very modest effect sizes (odd ratio less than 1.5). Therefore, any one variant has a minimal impact on the disease. These variants are also often difficult to identify using the GWAS approach, where many statistical tests often require stringent thresholds for declaring statistical significance.
- Rare genetic variants: Rare genetic single nucleotide polymorphisms (SNPs) are often defined as those variants with an allele frequency less than one percent. Research suggests that rare genetic variants often have greater overall genetic effects on psychiatric disorders [Frazer et al., 2009]. In fact some consistent genetic associations have been observed with rare genetic variants and several psychiatric disorders [Stefansson et al., 2008; Walsh et al., 2008]. Rare variants tend to be difficult to identify by the simple fact that few individuals have these variants, making large sample sizes necessary for variant identification.
- Copy number variants (CNVs): CNVs are alterations of the DNA of a genome that result in the cell having an abnormal number of copies of one or more sections of the DNA. As SNPs are alterations in a single nucleotide of genetic variants, CNVs are alterations of larger segments of genetic material. CNVs account for over 12% of the variation in the human genome [Stankiewicz and Lupski, 2010]. Some CNVs has been successfully identified for psychiatric disorders such as schizophrenia [Vrijenhoek et al., 2008], but these variants are also often rare, thereby limiting their identifiability.
- Incomplete genomic coverage: Although the costs for whole genomic sequencing are decreasing rapidly, to date there are few psychiatric genetic studies with whole genome sequencing on large numbers of individuals. As such, the GWAS and imputed data that are commonly being used may not often identify the true causative variants. In addition there may be no genotyped genetic variant that is in strong linkage disequilibrium with the actual causative variant. Therefore using current methodologies, causative variants often remain unidentified.
- Incomplete assessment of interactions: The single SNP approach is also limited in its ability to fully assess the interactions between multiple genetic variants. Seeing that many of the genetic variants that contribute to the pathogenesis of psychiatric disorders are likely interacting in a complex molecular network, it is likely that several genetic variants act together to “turn on” or “turn off” molecular pathways that increase or decrease disease risk. Therefore, by studying each genetic variant independently, the potential large impact that multiple genetic variants may have when acting in concert together is completely overlooked. In addition, it is highly likely that several of these genetic variants are impacted by known environmental risk factors (e.g., smoking, in-utero conditions) and when studied independently, these gene by environment interactions are also ignored.
- Heterogeneity of phenotypes: An inherent problem for complex disorders is the inevitable heterogeneity that exists within each diagnosis. For psychiatric disorders, this is exemplified by the subtype classifications for many disorders. For example, an individual diagnosed with bipolar disorder may have a sub-classification of bipolar 1, bipolar 2, bipolar no otherwise specified, or cyclothymic disorder. In genetic analyses, it is a constant deliberation whether to reduce the population to those individuals with a very specific, more homogeneous disorder or to include a larger number of individuals with a more heterogeneous disease. Regardless, the variation in outward manifestations of a given psychiatric disorder likely correlate with increased genetic heterogeneity, which will decrease our overall power to detect genetic associations.
In sum, psychiatric genetics is a complex field that has had some successes in identifying genetic variants that contribute to disease. Despite this, the majority of molecular determinants for most of these disorders remain to be discovered. We propose that an integrative “omics” approach offers promise to identify a portion of the “missing heritability” [Manolio et al., 2009] for these disorders.