GWAS have been a powerful tool for conducting unbiased scans of the genome to identify SNPs implicated in common, complex diseases, and represent an important advance compared to candidate gene studies. There is a wide gap, however, between the proportion of variance in disease susceptibility explained by the results of GWAS (usually between 1–10%) and the proportion of variance in disease susceptibility thought to be due to genetics based on heritability estimates (see Table 1) , which can be as high as 50% or greater. Furthermore, this is particularly true for neuropsychiatric disorders where GWAS have been less successful relative to studies of other common chronic aging-related diseases . Many reasons for this missing heritability have been proposed including the need for much larger sample sizes to detect additional SNPs of smaller effect that are yet to be found. In addition, rare variation (<1% frequency) and structural variation (e.g., copy number variants, insertions, deletions, inversions, and translocations), forms of variation that are not well-captured with most genotyping chips currently in use, likely account for some fraction of the unexplained genetic variability. Finally, low power to detect gene-gene interactions, inadequate examination of gene-environment interactions, phenotypic heterogeneity or imprecise phenotypic definition, and epigenomic alterations such as imprinting or parent-of-origin effects have also all been proposed as explanations for missing heritability .
Some of these explanations (e.g., phenotypic heterogeneity) may be particularly relevant for explaining missing heritability in neuropsychiatric disorders. Furthermore, with regard to mental illness in particular, the common disease, common variant hypothesis has been called into question, and it has been proposed that an alternative evolution-informed framework, characterized by the importance of gene-environment interactions and rare variants is more tenable for these types of disorders . To some extent, it is this issue of unexplained genetic variance that has hindered the use of genomic information from GWAS for the development of clinically useful predictive tests for common, complex (non-Mendelian) diseases. This, in turn, has limited the development of more targeted prevention strategies. We propose that one priority research area for leveraging genomics for disease prediction and prevention should be development of new strategies for finding and investigating the remaining heritability, especially in the area of neuropsychiatric diseases.