If you know the enemy and know yourself, you need not fear the result of a hundred battles. If you know yourself but not the enemy, for every victory gained you will also suffer a defeat. If you know neither the enemy nor yourself, you will succumb in every battle.

Sun Tzu, The Art of War

The greatest difficulty in finding treatments for schizophrenia is that we do not know the enemy well enough. Revealing the complex etiology and pathophysiology of schizophrenia has posed a considerable challenge for researchers, but improving technology is now enhancing our ability to use the wellspring of information present in the genome to help find these answers. It is more than three decades since the first development of genome sequencing technology,¹  and we have come to appreciate the intricate way in which variations in the genome can influence disease. Genetic research has provided insights into elucidating the pathophysiology of many diseases,^2,3  and also promises to improve clinical outcomes through personalized treatments and targeted therapeutics.^4–6  Studying the genetics of schizophrenia is important to discover genes and pathways that contribute to its development. The hope is that the symptoms of schizophrenia can be prevented or resolved by targeting therapeutics at these pathways. Still, treatment is most likely to be administered late in the development of the disorder, after diagnosable symptoms have already presented. By this time, the processes leading to the development of schizophrenia may have caused permanent changes; for example, alterations in brain morphology. Genetics can also help us to understand the underlying pathophysiology of the individual symptoms of schizophrenia, allowing for the development of targeted therapeutics to improve the lives of patients by treating symptoms after developmental pathways have become fixed. Whether to understand developmental processes or symptom pathophysiology, the study of the genetics of schizophrenia has great potential in helping us to understand the enemy, and hopefully, eventually, to conquer schizophrenia.

It is now understood that genes and environment work together to influence the development of disease. The power of genetics to enable us to understand a disease is dependent upon how much of the variance in liability is contributed by genes compared to other factors. It is also important to consider the manner in which genes affect phenotype. The heritability and genetic architecture of schizophrenia tell us how genetic information can be used to understand the disorder.

The contribution of genes in determining a given phenotype can be quantified by estimating heritability. Heritability is a mathematical expression of the amount of variance in phenotype that is explained by genetic variation. This does not measure how much phenotypic variation is caused by genes; rather, it reflects the relative contribution of genetic vs. non-genetic factors in determining phenotype. Heritability is estimated by comparing the liability of developing a trait (schizophrenia, for example) between related and unrelated individuals.^7,8  Twin studies have been invaluable for estimating heritability, as it is easier to differentiate between genetics and shared vs. differential environment in such studies.⁹  The concordance in phenotype between monozygotic (MZ) and dizygotic (DZ) twins gives a measure of the correlation between genotypic variation and presence of a trait. MZ concordance rates for schizophrenia have been reported between 41% and 65%, with DZ concordance ranging from 0% to 28%.^10,11  Since DZ twins have approximately half the genetic variance of unrelated individuals, and MZ twins have identical genomes, heritability can be crudely calculated as twice the difference in concordance (r) between MZ and DZ twins (see eqn 1.1).⁸ 

Hence, the heritability for schizophrenia estimated from twin studies is 81% (95% confidence interval 73–90%).^10–13  A limitation of twin studies is that subjects are usually recruited from restricted environmental settings, typically from within the same hospital. Heritability estimates using family data from national records are lower, at 64–67%.^12,14  This difference may be due to increased variance in environment and diagnostic interpretations when using subjects from national records. Either way, schizophrenia is clearly one of the most heritable neuropsychiatric disorders, demonstrating that genes have a large role to play.

The high estimated heritability for schizophrenia indicates that the genome contains information explaining much of the underlying pathophysiology of the disorder. So far, all the variants taken together from current genome-wide association study (GWAS) results have been calculated to account for 20–40% of the variation in liability for schizophrenia.^15–18  Even using lower family-based estimates for comparison, current results do not account for all of the predicted heritability. The term “missing heritability” was coined to describe this discrepancy between the proportion of phenotypic variation explained by results from genetic studies and the total estimated heritability.¹⁹  Current evidence suggests that we may yet find a large proportion of this missing heritability within the genome. Part of the missing heritability may also be due to epigenetic DNA and chromatin modifications that alter gene expression without changing DNA sequence. Increased sample sizes and more complete coverage of variants with improved genotyping technologies have already uncovered many new significant schizophrenia-associated genetic loci. We can be optimistic that continuing efforts in interrogating the human genome will reveal ever increasing numbers of causal variants. This information promises to provide key insights into the mechanisms underlying the development of schizophrenia.

Genetic studies have now discovered enough associated risk variants to give an empirical view of the genetic architecture of schizophrenia.^20–22  Schizophrenia is a complex, highly polygenic disorder with multiple variants conferring risk. Numerous variants with population frequencies >1% have been associated with schizophrenia. Alongside these common variants, rare variants with frequencies <0.1%, resulting from de novo mutations and large-effect structural variations, are implicated in the disorder. The effect size of schizophrenia associated variants is inversely proportional to their population frequency. Effect size encompasses the idea of penetrance. Penetrance reflects the amount an individual variant contributes to a phenotype, with completely penetrant variants guaranteeing the presentation of a trait, and incomplete and low penetrant variants only incrementally increasing the probability of possessing that trait.

The allelic spectrum of schizophrenia risk is depicted in Figure 1.1. Multiple low effect common variants with odds ratios (ORs) typically <1.3 and moderate to high effect, but still incompletely penetrant rare variants contribute to risk. Risk variants can combine additively where each locus adds/subtracts a certain amount of risk, or multiplicatively where a certain number or arrangement of loci must be present to reach a threshold to increase risk. A purely additive model of gene interaction does not explain the genetics of schizophrenia.^23–25  Therefore, the genetic architecture of schizophrenia probably involves thousands of risk variants which interact multiplicatively and are highly susceptible to genetic background, pleiotropy, and, of course, environmental effects. Unaffected individuals probably carry manageable numbers of risk variants, while the probability of developing schizophrenia rises sharply for individuals with a high burden of risk alleles. While this picture suggests that common variants are unlikely to be essential alone, the multiplicative nature of their effects means that single risk variants can still exert biologically meaningful effects, depending on the genetic background.²⁴  Hence, associated common variants still warrant functional investigation alongside higher risk rare variants. The picture so far suggests that understanding the genetic causes of schizophrenia will involve understanding not only the functional relevance of individual genes, but also how multiple genes interact and converge on molecular pathways to affect behaviour.

New technologies are constantly improving our ability to read the human genome and detect genetic variation between individuals. Genomes vary in many ways, from single nucleotide polymorphisms (SNPs) to more severe structural variations such as copy number variations (CNVs). SNPs are the most common type of variation.²⁶  Structural variants have a greater potential to cause disruptions in genes simply due to their size, and hence large-effect rare variants tend to be of the structural type; however, other variants can still have large effects.²⁷  A simple schematic of the types of variation is given in Figure 1.2. Different techniques and technologies are better for detecting specific types of variation, so study design and genotyping methods must be tailored to the type of variation relevant to the specific research question.

Sequencing remains the gold standard to capture all of the variation in the genome. Next-generation sequencing (NGS) technology has greatly reduced the cost of sequencing and is now standard practice.^6,28  NGS sequences millions of DNA fragments in parallel, with bases being identified optically in real time using “sequencing by synthesis” chemistry, greatly reducing time and costs.^29–31  Third-generation sequencing and nanopore based technologies are on the horizon; these can sequence single DNA molecules without the need for amplification or cyclic-sequencing steps.^32–36  Sequencing is still too expensive for large samples, so microarray technology remains the most used genotyping method in GWASs.³⁷  Modern microarrays can now simultaneously interrogate up to 1 million SNPs, and are capable of detecting CNVs and microsatellites; however, inversions and translocations can only be detected by comparison of fully sequenced genomes.^38–40  Only a fraction of SNPs can be interrogated on a microarray; however, un-interrogated SNPs can be mathematically predicted from reference genome data by imputation, greatly increasing coverage.^41–46  High coverage is crucial for detecting disease variants as it is unlikely that the variants on an array are causal; rather, they are in linkage disequilibrium with true causal variants. Cheaper sequencing technologies, improved arrays, and better reference information for imputation will continue to increase our ability to find disease-associated variants.

Study design can greatly affect the ability to find genotype–phenotype associations. There are two major study designs used in human genetics: case–control and pedigree-based. Simpler case–control studies are better for finding associations with low effect size, but cannot discriminate between inherited and de novo variations.^47–49  More complicated family-based designs can be used to evaluate linkage (co-segregation of genotypes and phenotypes from parents to offspring), test for associations, and identify de novo variants.^49,50  Subject choice is important as families with a history of disease (multiplex pedigrees) may be enriched for rare causal variants, whereas affected subjects whose families have no history of disease (simplex pedigrees) may be enriched for de novo variants. It is also important to consider how data are analyzed. False discovery rate procedures to correct for multiple testing, test-replication designs, and pathway analysis for enrichment of functionally-related genes are all clever ways to increase statistical power without relying on massive sample sizes.^49–53  Carefully considered study designs combined with constantly improving technologies are already generating results in the search for causal variants for schizophrenia, and we can expect continued progress in schizophrenia genetics.

Even in the late 2000s there was a worry that genome-wide screens were finding no true causal variants for schizophrenia.^54,55  While early linkage studies were beginning to find significant loci, many of the most interesting findings were not replicated, and candidate genes tested from significant loci yielded no associations.^56–59  Past GWASs yielded few strong results due to lack of power;^55,60,61  however, newer GWASs and mega-analyses with combined sample sizes in the tens of thousands are now finding numerous significant associations.^17,60,62,63  Furthermore, improving genotyping technology and analysis techniques are making it possible to determine the role of rare structural variation and de novo mutations in schizophrenia, and facilitate the identification of rare associated CNVs.^64–67  Many schizophrenia risk variants are beginning to show biological relevance and potential as drug targets. There are now too many associated loci to mention in any adequate detail in this section; however, some of the more interesting results are highlighted.

Common SNPs account for a large amount of the variance in liability for schizophrenia. Using statistical models, it was estimated in 2012 that 23% of the variation in liability for schizophrenia is accounted for by common SNPs.¹⁶  Just 1 year later, that estimate had increased to at least 32%, and this number should continue to climb as better-powered studies discover more significant associations.¹⁷  A few of the more robust and interesting GWAS findings are summarized in Table 1.1. Some of the more interesting themes arising from genetic studies in terms of drug discovery are detailed below.

Receptors represent obvious potential therapeutic targets. The DRD1 gene encoding the D1 dopamine receptor gained strong epidemiologic credibility from early genetic studies.⁶⁰  While implicating the dopamine system was not a new finding, this is an example showing that genetic studies were corroborating established hypotheses. Common SNPs in the receptor genes CHRNA7 and GRM3 were also associated with schizophrenia, but, as was typical with early candidate gene studies, many studies also reported no associations.^60,68  Nevertheless, concordant evidence supported a role for these receptors.^69,70  CHRNA7 encodes a subunit of the ionotropic α-7 nicotinic acetylcholine receptor (nAChR). This receptor seems to function mainly to modulate neurotransmitter release in the striatum.^71,72  There is evidence that α-7 nAChR agonists have efficacy in improving cognitive deficits in schizophrenia, and may be useful in combination with antipsychotics.^73–76  Additionally, rare variants in CHRNA7 show strong associations with schizophrenia.^69,77  GRM3 encodes the mGluR3 subunit of the metabotropic glutamate receptor (mGluR). Allosteric and orthosteric modulators of mGluR2/3 are available,^78,79  and evidence from animals and early clinical trials is beginning to show that some of these agonists may have efficacy in treating the positive and negative symptoms of schizophrenia.^70,80–83  These examples show how genetic research with concordant biological evidence allows us to tap new sources with therapeutic potential.

Associations within an exceptionally complex genomic region on chromosome 6 known as the major histocompatibility complex (MHC) are some of the most robust and consistent findings for schizophrenia.^{17,18,63,84–86}  This region contains hundreds of genes in high linkage disequilibrium, thus making it difficult to identify specific genes underlying associated loci.^49,87,88  Nevertheless, examining the general role of the MHC in immune function, autoimmunity, inflammation, and infection in relation to schizophrenia suggests intriguing new directions in schizophrenia research.⁸⁹  Animal and in vitro studies are revealing a role for MHC molecules in neurodevelopment, neuronal and synaptic plasticity, learning, memory, and behavior.^90–94  Furthermore, MHC molecule expression is altered in schizophrenia, and risk loci within the MHC have been associated with functional effects on brain morphology and cognition in humans.^95–100  MHC associated autoimmune disorders, infections with certain pathogenic microbes, and prenatal maternal immune activation have been associated with increased schizophrenia risk;^101–111  in addition, neuroinflammation and increased inflammatory markers are seen in patients with the disorder.^112–120  Systematic reviews and meta-analyses show that anti-inflammatory drugs given in combination with antipsychotics decrease the severity of schizophrenia symptoms.^121–123  A complete review of the intricate relationship between the immune system and the central nervous system as it relates to the MHC and schizophrenia is beyond the scope of this chapter; however, it is clear that associations in the MHC can inform our understanding. Additionally, it is likely that therapeutic strategies involving the immune system will improve the treatment of schizophrenia.

Genetic studies have implicated a number of protein kinases in schizophrenia. Associated SNPs in AKT1 were found in diverse populations with few negative reports,^124–131  and a strong association was recently found in AKT3.¹⁷  The encoded protein AKT, also called protein kinase B, is activated downstream of glutamate signaling in neurons, and mediates phosphoinositide (PI)3 kinase-derived signaling.^132,133  Additionally, AKT is upstream of glycogen synthase kinase (GSK)3β, which mediates its effects. SNPs in TAOK2 and MAP2K7 have also been associated with schizophrenia.^134,135  Thousand-and-one-amino acid 2 kinase (TAOK2) regulates cortical neuronal morphology.¹³⁶  Mitogen activated protein kinase kinase (MAP2K)7 regulates axon development in the cortex, and knocking out MAP2K7 in mice results in schizophrenia-like behavioral deficits.^135,137  Furthermore, TAOK2 and MAP2K7 are both involved in the c-Jun N-terminal kinase (JNK) signaling pathway, suggesting a role for this pathway in schizophrenia.^137–139  Lastly, linkage regions 8p21–12 and 2q33.3–34, as well as SNP and microsatellite variants in ERBB4 and NRG1, have been associated with schizophrenia.^{56,57,60,140,141}  The encoded protein neuregulin (NRG)1 activates receptor tyrosine-protein kinase erbB-4 (ERBB4) to initiate downstream signaling involving JNK, extracellular signal-regulated kinase (ERK) and PI3 kinase pathways. NRG1 is involved in neurodevelopment, and NRG1/ERBB4 signaling is implicated in glutamatergic, γ-aminobutyric acid (GABA)ergic and dopaminergic neurotransmission.¹⁴²  NRG1 and ERBB4 function in relation to schizophrenia has been extensively studied in mouse models.¹⁴³  Rare variants in TAOK2 and ERBB4 have also been associated with schizophrenia.^67,144–149  Each of these kinases, and their related pathways, is an excellent potential target for therapeutic intervention. The role of these kinases in schizophrenia will be elucidated with further research.

One of the newest findings from GWASs implicates genes encoding l-type calcium channel subunits CACNA1C and CACNB2 in schizophrenia. These channels play a role in learning, memory, and synaptic plasticity, and have also been associated with autism, bipolar disorder, and the calcium channelopathies Brugada and Timothy syndromes.^150–156  Many approved medications act on calcium channels; for example, some antipsychotics (e.g., pimozide) and adjuvants for non-responders in schizophrenia and bipolar disorder (e.g., verapamil and nifedipine).^157,158  Hence, these associations not only suggest potential mechanisms for the etiology and pathophysiology of many neuropsychiatric disorders, but also provide hypotheses for quick clinical translation through the repurposing of approved medications.

One of the strongest recent associations is the MIR137 gene coding the microRNA miR-137. This is particularly notable as miR-137 functions to regulate multiple genes by binding target sites present on mRNA.¹⁵⁹  It is highly expressed in the brain, and is an important regulator of neurogenesis and neuronal maturation.^160–163  Genes with predicted miR-137 binding sites were enriched for lower p-values;¹⁷  furthermore, many predicted and confirmed miR-137-regulated targets reached genome-wide significance, including HLA-DQA1, CACNA1C, CACNB2, ZEB2, CSMD1, MAD1L1, DPYD, TCF4, and many others.^{17,63,164,165}  Hence, miR-137 stands at the top of multiple potential schizophrenia-associated pathways. In addition to miR137, multiple regions containing long intergenic non-coding RNAs (lincRNAs) recently reached genome-wide significance.¹⁷  The function of these lincRNAs is not well understood, but they may have roles in epigenetic regulation and development.¹⁶⁶  These findings provide a myriad exciting new research directions to understand schizophrenia and find new therapeutic targets.

Many rare but potent structural variants have been discovered to have a role in a small proportion of schizophrenia cases; however, these variants tend to be non-specific and associate with multiple disorders, such as bipolar disorder, autism, intellectual disability, epilepsy, and others.⁴⁹  Finding associated rare variants presents a unique set of challenges in genetic research. Technological advances have driven much of the search for rare structural variants.^39,40,64  With decreasing sequencing costs, and new microarrays capable of detecting CNVs, searching for structural variants in large epidemiological studies is becoming more feasible. Table 1.2 provides a summary of some of the structural variations associated with schizophrenia, most likely representing the “low-hanging fruit” of rare variations. These tend to be large, centering on structural variation hotspots.¹⁶⁷  Many of these variants span multiple genes, so further research is needed to narrow the focus and understand their functional roles in disease. Nevertheless, some more specific directions have emerged from the study of rare variation. More rare variations will be found as new technology for detecting structural variation is applied to ever larger samples.

DISC1 was discovered at the breakpoint of a chromosomal translocation (t1q42.1; 11q14.3) in a Scottish pedigree which presents with severe psychiatric disorders, including schizophrenia, depression, and bipolar disorder.^168,169  While this translocation was not found outside this family, additional associations with common and rare variants in DISC1 were found.^60,170–173  DISC1 encodes a multifunctional scaffolding protein involved in regulating embryonic and adult neurogenesis, and neuronal proliferation, differentiation and migration.^{170,174–176}  Disrupting DISC1 in animal models causes schizophrenia-like behavioral deficits.^177–180  Indeed, converging evidence from genetics and functional studies strongly supports a role for DISC1 in schizophrenia.¹⁷⁶  DISC1 interacts with many potential target proteins that are involved in synaptic function, and neurodevelopmental, cytoskeletal, and centrosomal pathways, some of which are also associated with schizophrenia (e.g., AKT, DPYSL2, GSK3β, PDE4, and TNIK).^{173,181–186}  While DISC1 itself is not the easiest target for therapeutics, it is an example of hypothesis generation from genetic research greatly contributing to our understanding of schizophrenia.

Rare CNVs in NRXN1, including a de novo instance, have been associated with schizophrenia.^{67,77,187–190}  CNVs in NRXN1 are also associated with autism.^191,192  NRXN1 encodes members of the neurexin superfamily of proteins, which are presynaptic cell adhesion molecules which form heterotypic intercellular junctions with neurologin across synapses.¹⁹³  NRXN1 is one of the largest known human genes, and is regulated through alternative splicing at its six splice sites yielding numerous isoforms, each with unique binding affinities. Neurexins are thought to be involved in synapse and neuronal maturation and have been implicated in neurodevelopmental pathways.^194,195  Furthermore, neurexins possess an intracellular PDZ domain which can interact with many presynaptic proteins.¹⁹⁴  The potential role of NRXN1 in schizophrenia remains to be elucidated; hence, NRXN1 represents an interesting future direction in schizophrenia research.

Duplications in the VIPR2 locus at 7q36.2 are strongly associated with schizophrenia,^77,196  suggesting changes in gene dosage may affect the disorder. Low copy repeats at the VIPR2 gene may predispose it to structural variations (see Figure 1.2).¹⁹⁷  VIPR2 encodes the type II vasoactive intestinal peptide G-protein coupled receptor (also called VPAC2), which is coupled to adenylate cyclase activity and expressed in the cerebral cortex and thalamus.¹⁹⁸  Changes in gene dosage causing alterations in VIPR2 signaling could hypothetically be associated with schizophrenia etiology or symptoms; hence, VIPR2 represents a good potential target for therapeutics. Future research into VIPR2 gene and protein function will reveal its relationship to schizophrenia.

Sequencing allows researchers to analyze the genome to a level of detail that is otherwise unattainable; however, high costs continue to limit the size of sequencing studies.¹⁹⁹  While insufficient statistical power is still precluding the ability to detect significant loci in these studies, clever analysis techniques are revealing interesting trends for rare variation. Evaluating the burden of structural variation in cases compared to controls tests a multigenic hypothesis in which many rare but different disruptions contribute to disease risk. Multiple studies report an increased burden of CNVs in schizophrenia.^67,200–202  Additionally, a role for de novo mutations in schizophrenia has been reported in a number of studies.^{200,203–207}  While the evidence for increased rates of de novo variation is mixed, de novo variants disproportionately disrupt genes in schizophrenia populations, suggesting a functional role. Pathway analysis has been used to great effect in schizophrenia sequencing studies.^67,148,208  Genes within rare variants are significantly enriched in functionally-related gene sets, mostly composed of synaptic proteins; specifically, the voltage gated calcium ion channel, genes within the activity-regulated cytoskeleton-associated (ARC) protein signaling complex, N-methyl-d-aspartate (NMDA) receptor complexes, and glutamatergic postsynaptic proteins.^201,203  A caveat of pathway analysis is that there are many potential biases that result from incomplete data sets, differences in gene size, or multiple pathway membership.^20,52,53,209  Nevertheless, the results for schizophrenia appear to be etiologically plausible. The future of sequencing in genetic studies looks promising; however, caution should be taken as handling the massive amounts of data from fully sequenced genomes presents challenges for statistical analysis. The falling cost of sequencing will allow these technologies to be applied to the large samples needed to detect specific variants.

Moving forward with studies interrogating the genome for schizophrenia-associated risk variants, a number of important concepts have emerged. The lack of results from early studies was primarily due to a lack of statistical power, and increasing sample size is the easiest remedy for this problem.^16,17,21,55  Working out the logistics to handle samples in the tens of thousands and beyond will be the challenge for future researchers. Evidence suggests that heterogeneity across diverse groups may be low for schizophrenia;^210,211  hence, planned mega-analyses across world populations may be worth pursuing.²⁰  As sample size increases, the number of loci reaching genome-wide significance will climb. In addition, it is important to understand that many associated loci may not be truly causal, but simply in linkage disequilibrium with causal variants. Improved microarrays and reference data for imputation, as well as cheaper sequencing technologies will improve our ability to find true causal loci, especially rare variants that are less likely to be included on commercial microarrays.⁶  Analysis techniques are also improving. The increasing body of gene function literature, improved categorization and annotation of functions, and better algorithms will increase the utility of pathway analysis as a legitimate tool for supporting hypotheses and corroborating evidence of associations.²¹²  Interestingly, many schizophrenia-associated loci are also implicated in other disorders, such as autism spectrum disorders and bipolar disorder.^20,213  Furthermore, schizophrenia associations are enriched in functionally annotated genes.²¹⁴  While complicating the picture for schizophrenia genetics, this information can be used beneficially to estimate the false discovery rate and improve power using purely statistical methods.^212,215,216  These analysis techniques can be applied to existing data sets as well as new studies to find associations that were previously hidden by statistical noise. Increasing sample size, improving technology and statistical techniques, and careful study design will facilitate the search for schizophrenia risk variants in the future.

Long-held distinctions in psychiatric nosology and diagnostic manuals have attempted to separate psychiatric illness into defined categories having discrete boundaries.^217,218  Yet the diagnostic taxonomy of schizophrenia has always necessitated the use of multidimensional criteria, as many core symptoms transcend the boundaries of multiple disorders.^219–223  The picture emerging from genetic studies suggests that a paradigm shift may be needed, moving away from discrete classifications. Certainly, one of the recurrent themes from Section 1.3 is the non-specific nature of associated loci. Many of the rare variants in Table 1.2 are associated with developmental disorders, intellectual disability, and autism. Furthermore, multiple schizophrenia-associated SNPs show cross-disorder associations with autism spectrum disorder, attention-deficit hyperactivity disorder, bipolar disorder, and major depressive disorder.²²⁴  For example, SNPs in CACNA1C, ZNF804A, ITH3-ITH4, and ANK3 were also associated with bipolar disorder.^18,20,213 

Phenomena such as epistasis, variable expressivity, pleiotropy, or gene–environment interactions could explain how different disease states arise from individual genetic variants. It could be the case that these variants are mediating the same symptoms in different disorders, and are susceptible to effects such as epistasis and environmental interactions which modify the severity of their presentation. Hence, rather than approaching schizophrenia as a unified disorder, it may be beneficial from a gene discovery perspective to focus on individual symptom domains that are likely to be more proximal to the underlying neural substrates, and therefore genes, than the disorder as a whole.

Segregating analysis by symptom domains is related to the idea of stratifying genetic analysis based on endophenotypes. Endophenotypes can be defined as a subset of biomarkers which are heritable and increase the risk of an illness.^225,226  These can be any type of trait; for example, neuroimaging, electrophysiological, and cognitive variables are all considered endophenotypes for schizophrenia.^227–232  There are a number of advantages to using endophenotypes: they are simpler and closer to the gene level of action, so associated variants are more likely to show larger effect sizes; they allow for easier stratification of populations and quantitative ranking within a diagnostic category to increase statistical power by decreasing heterogeneity; and they are more easily translatable to animal models, so it is easier to perform functional studies of genes.²³³ 

Understanding individual symptom domains or endophenotypes also has consequences for drug development.²³⁴  Many schizophrenia risk variants, particularly rare variants, affect neurodevelopmental pathways that affect brain structure.^20,231,235  Unless intervention occurs pre-emptively during development, medications are unlikely to be useful in these pathways if the changes in brain structure underlie schizophrenia symptoms. Conversely, much of the functional disability associated with schizophrenia is due to cognitive impairment.^236–238  Therefore, understanding the genetic and molecular underpinnings of cognitive symptoms is important for developing therapeutics that will vastly improve the ability of schizophrenic patients to function.^239,240  While overlap is almost guaranteed between domains such as brain structure and cognitive impairment, consideration of factors that affect one but not the other is important for informing efforts into drug discovery. Endophenotypes are already a useful tool for functional analysis of schizophrenia risk variants.^{95,230,231,241,242}  The use of endophenotypes or symptom domains to refine future GWAS analyses will aid the search for new risk variants and facilitate our understanding of the genetic mechanisms underlying schizophrenia.

Genetics is an excellent tool for directing research to improve our understanding of schizophrenia. The emerging story from genetics has moved schizophrenia research in interesting new directions. It has also forced us to abandon some of our past expectations. A longstanding issue in schizophrenia is the reliance on clinical phenomenology for diagnosis.²⁴³  Genetic research was expected to be a source of biologically valid markers to improve diagnosis. Sadly, the picture from genetic research shows that this hope may go unfulfilled. The genetic architecture of schizophrenia, consisting of thousands of low effect variants and variants demonstrating pleiotropy for other illnesses, demonstrates that genetic variants are, so far, unlikely to be useful as diagnostic tools.²⁴ 

While great progress has been made in accounting for the “missing” heritability in schizophrenia, it should be noted that some of the predicted 81% may remain missing. Gene–environment interactions tend to be attributed to heritability in epidemiological studies.²⁴⁴  MZ twins, sharing nearly 100% of their genes, have a concordance of ∼50%, emphasizing the importance of environmental input. Furthermore, unusual genomic effects, such as over-dominance,²⁴⁵  and epigenetic factors have been implicated in contributing to the heritability of schizophrenia.^246–248  Gene–environment interactions can be incorporated into genetic research with careful phenotyping, and would greatly improve our understanding of schizophrenia risk.²⁴³  However, as it stands, the environment is largely ignored. Finally, it should be emphasized that genetics, above all, is a tool for hypothesis generation. Barring unusually strong Mendelian associations, the results of genetic research give probabilistic associations of varying credibility. Many strongly associated variants do not fall within genes; rather, they are located in intergenic regions or within introns, and have unknown functional relevance.^17,60,62,63  False positives are always a danger, but rigorous avoidance of potential false positives can also eliminate evidence of the most interesting results.^20,50,51  It is important to use discretion when moving forward with results, and as with all science, multiple lines of convergent evidence are desirable.

The use of genetic information for drug discovery is still in its early development. Progress in schizophrenia research was slow early in the past decade, but we are now finally approaching sufficient sample sizes to accelerate the discovery of risk variants. Genetic discoveries have associated a plethora of genes with schizophrenia, and highlighted the complex, heterogeneous nature of the disorder. Further research to replicate important findings and functionally characterize risk genes is important to understand the underlying processes involved in schizophrenia. Validation of variants and a functional understanding of how specific mutations lead to the development and presentation of the disorder are essential for novel drug discovery.

While there is no risk variant yet discovered that has an effect on a significant number of schizophrenia patients, the major discoveries so far appear to converge on clinically relevant pathways. These findings have advanced our understanding of the etiology and pathophysiology of schizophrenia. Given the large overlap between risk variants for schizophrenia and other psychiatric disorders, priority should be directed to pathways which underlie specific symptom domains that cross current diagnostic categories. Genetic research will continue to find new associated variants and suggest new directions for understanding schizophrenia. The challenge for future researchers will be to weave together the intricate picture from genes to proteins to pathways, and finally to the generation of novel therapeutics.

Coverage: the estimated proportion of the genome that can be captured by SNPs interrogated on an array at a preset correlation threshold.

de novo variant: a causal variant that is the result of a mutation occurring for the first time within a subject; i.e., it was not inherited.

Epistasis: when the effects of one gene depend on one or more additional modifier genes (i.e., the genetic background).

Gene–environment interaction: when one gene exerts differential effects on a phenotype when exposed to different environments.

Odds ratio (OR): the ratio of the probability of an event occurring given exposure over the probability of an event occurring given no exposure. Exposure can be to anything; however, in genetic studies it usually refers to the presence of a certain variant. The OR in the context of genetics is also called the genotypic relative risk.

Over-dominance: a genetic condition in which the phenotype of the heterozygote lies outside the phenotypic range of the homozygotes at any allele.

Pleiotropy: when one gene exerts effects on multiple apparently unrelated phenotypes.

Population prevalence rate: the proportion of a population with a given trait over a given time.

Statistical power: the probability that a statistical test will reject the null hypothesis when the alternative hypothesis is true; i.e., the probability of not committing a type II error. Effectively, it is a measure of the ability of a study to find true significant results.

Variable expressivity: when individuals with the same genotype express a phenotype to a different degree from each other.