Genetic architectures of psychiatric disorders

Nature Reviews Genetics | AOP, published online 10 July 2012; doi:10.1038/nrg3240
R E V I E W S
D I S E A S E M E C H A N I S M S
Genetic architectures of psychiatric
disorders: the emerging picture and
its implications
Patrick F. Sullivan1, Mark J. Daly2 and Michael O'Donovan3
Abstract | Psychiatric disorders are among the most intractable enigmas in medicine.
In the past 5 years, there has been unprecedented progress on the genetics of many of
these conditions. In this Review, we discuss the genetics of nine cardinal psychiatric
disorders (namely, Alzheimer's disease, attention-deficit hyperactivity disorder, alcohol
dependence, anorexia nervosa, autism spectrum disorder, bipolar disorder, major
depressive disorder, nicotine dependence and schizophrenia). Empirical approaches have
yielded new hypotheses about aetiology and now provide data on the often debated
genetic architectures of these conditions, which have implications for future research
strategies. Further study using a balanced portfolio of methods to assess multiple forms
of genetic variation is likely to yield many additional new findings.
A core set of psychiatric conditions -- madness, mania association studies (GWASs) and structural variation studies,
Pellagra
A disease caused by niacin
and melancholia -- has been perplexing for millennia. although studies of uncommon or rare exonic variation
(vitaminB ) deficiency that has
3
Although mortality is increased for many psychiatric are likely to have a prominent role in the next few years.
prominent neuropsychiatric
disorders1, their major impact is on morbidity: psychi- These results meet community standards in human
symptoms.
atric disorders account for around one-third of disability genetics for significance and replication6. Although these
Neurosyphilis
worldwide2 and cause enormous personal and societal findings often appear in high-profile journals, sentiments
Infection of the central nervous
burdens3.
such as `genetics has failed in psychiatry' or `there are no
system by Treponemapallidum
In the past century, considerable efforts towards genes for psychiatric disorders' are still heard. A review of
(which is the spirochete that
understanding the nature of psychiatric disorders have psychiatric genetics is thus particularly opportune.
causes syphilis) and often
been undertaken. There have been successes, and a few
Over 300 psychiatric disorders have been described,
causes prominent
neuropsychiatric symptoms.
diseases (for example, pellagra4 and neurosyphilis5) with and nine are covered in this Review. The conditions
prominent psychiatric manifestations that were previ- selected are all psychiatric disorders that have been sub-
ously prevalent are now rare in many parts of the world. jected to intensive genetic study and for which genome-
1Departments of Genetics
These few triumphs stand in contrast to decades of frus- wide results (usually GWASs and structural variation
and Psychiatry, CB# 7264,
tration and occasional notoriety, when highly publicized but also genome-wide linkage and resequencing) have
5097 Genomic Medicine,
and plausible findings failed to be replicated. Indeed, been obtained. The disorders and their abbreviations are
University of North Carolina at
most psychiatric disorders have been intractable to defined in TABLE1 and Supplementary informationS1
Chapel Hill, North Carolina
27599-27264, USA.
approaches that were fruitful in other areas of medicine. (table), and the heritabilites and lifetime prevalences are
2Department of Genetics,
Thus, psychiatric syndromes are general y referred to as depicted in FIG.1a. Intel ectual disability could have been
Harvard University,
`disorders' (which are illnesses that disrupt normal func- included, but the voluminous literature on this topic has
Cambridge, Massachusetts,
tion), and only a few are referred to as `diseases' (which been thoroughly reviewed7-9. Studies of other psychiat-
USA.
3MRC Centre for
are disorders with a known pathophysiology or structural ric disorders are in progress, but the published data are
Neuropsychiatric Genetics,
pathology). An obvious goal of psychiatric research is to few (for example, in the cases of obsessive-compulsive
Cardiff University,
convert idiopathic disorders into pathophysiologically disorder, Tourette's syndrome and panic disorder).
Cardiff CF14 4XN, UK.
defined diseases.
The genetic dissection of complex traits has been
Correspondence to P.F.S.
Since 2007, numerous robust and replicable genetic frequently reviewed6,10-12, so here we provide a brief
e-mail: pfsulliv@med.unc.edu
doi:10.1038/nrg3240
findings have been reported for psychiatric disorders. overview of the approaches and study design considera-
Published online 10 July 2012
These advances have mostly been through genome-wide tions in BOX1. Advances in genetics are often yoked to
NATURE REVIEWS | GENETICS
ADVANCE ONLINE PUBLICATION | 1
(c) 2012 Macmillan Publishers Limited. All rights reserved

---

R E V I E W S
Table 1 | Defining features of nine psychiatric disorders*
Name
Life
Heritability Essential characteristics
Notable feature
prevalence
Alzheimer's disease 0.132
0.58
Dementia, defining
Of the top ten causes of
neuropathology
death in the United States,
Alzheimer's disease alone has
increasing mortality
Attention-deficit
0.053
0.75
Persistent inattention,
Costs estimated at
hyperactivity
hyperactivity, impulsivity
~$US100 x 109 per year
disorder (ADHD)
Alcohol
0.178
0.57
Persistent ethanol use despite
Most expensive psychiatric
dependence (ALC)
tolerance, withdrawal, dysfunction
disorder (total costs exceed
US$225 x 109 per year)
Anorexia nervosa
0.006
0.56
Dangerously low weight from
Notably high standardized
self-starvation
mortality ratio
Autism spectrum
0.001
0.80
Markedly abnormal social
Huge range of function, from
disorder (ASD)
interaction and communication
people requiring complete
beginning before age 3
daily care to exceptional
occupational achievement
Bipolar disorder
0.007
0.75
Manic-depressive illness, episodes
As a group, nearly as
(BIP)
of mania, usually with major
disabling as schizophrenia
depressive disorder
Major depressive
0.130
0.37
Unipolar depression, marked and
Ranks number one in the
Genome-wide association
disorder (MDD)
persistent dysphoria with physical
burden of disease in the
studies
and cognitive symptoms
world
(GWASs). Unbiased genome
screens of unrelated cases and
Nicotine
0.240
0.67
Persistent nicotine use with physical Major preventable risk factor
appropriately matched
dependence (NIC)
dependence (usually cigarettes)
for many diseases
controls or parent-affected
Schizophrenia
0.004
0.81
Long-standing delusions and
Life expectancy decreased
child trios. The dominant
(SCZ)
hallucinations
by 12-15years
technology has been individual
genotyping using highly
*Most of these definitions are made more restrictive by requiring persistence over time (for example, the criteria for SCZ require
multiplexed SNP arrays.
6months of symptoms), substantial impairment and presence across multiple different contexts. See Supplementary information
S1 (table) for more detail. Additional sources are REFS1,2,181-183).
Structural variation
A genomic alteration that
changes the number of copies
technological advancements. Major approaches that have early-onset familial Alzheimer's disease27. These loci have
or the arrangement of the
been informative in psychiatric genetics include assess- atypical y large effect sizes, thereby facilitating identifica-
genome. Copy number variants
ment of: structural variation through karyotyping, array- tion using `past generation' technologies, such as candi-
are one type of structural
variation.
based methods and high-throughput sequencing13-16 date gene association and genome-wide linkage studies
(TABLE2 and Supplementary information S2 (table)); (Supplementary information S2 (table)). Treatments
Genome-wide linkage
GWASs using highly multiplexed SNP arrays and, for Alzheimer's disease based on these findings have
A type of unbiased genome
potential y, high-throughput sequencing12,17-19 (TABLE3); been developed and are undergoing testing. Rare struc-
screen based on multiplex
and high-throughput sequencing to uncover rare vari- tural variation duplications containing APP have been
pedigrees. Genotyping
approaches have included
ants of fairly strong effect (perhaps arising denovo)20,21. associated with Alzheimer's disease28,29. Small-exome
restriction fragment length
Genome-wide linkage and hypothesis-driven candidate sequencing studies of Alzheimer's disease have been pub-
polymorphisms, microsatellites
gene association studies have also been conducted but, as lished30,31, and larger studies are in progress and should
and SNP arrays. After
in many areas of biomedicine, with uncertain yield22-26.
provide a more nuanced understanding of the role of rare
adjustment for multiple
comparisons, the signal is the
In this Review, we first summarize the literature for the exonic mutations in the pathogenesis of the disease.
co-segregation of a genotype
nine disorders listed in TABLE1 with a particular emphasis
with a disease phenotype
on the findings that appear to meet community standards Common variation. In the early 1990s, apolipoprotein E
within the pedigrees.
for replication in human genetics (that is, robustly sig- (APOE) was identified by candidate gene association as
Karyotyping
nificant with consistent effects across samples)6. We then a susceptibility gene for late-onset Alzheimer's disease, in
Determination of the
highlight new hypotheses that have emerged across the no small part owing to its unusual y large effect size27,32
microscopic appearance and
al elic spectrum, including denovo and rare exonic muta- (TABLE3). In 2009, GWASs from two large consortia33,34
gross changes in chromosomal
tions, rare structural variation and common variation implicated three novel loci, and six additional loci were
content and structure of a cell.
from GWASs. Crucial y, these results provide empirical identified in 2011 (REFS35,36). Full meta-analyses are
Meta-analyses
insights into the genetic architectures of these disorders: keenly awaited, but the ten loci that have been identified
These are methods for
data that are essential for guiding future work in this area. to date account for ~20% of the total variation in risk or
summarizing and combining
~33% of the risk attributable to genetic effects, with the
results across multiple studies
Alzheimer's disease
major contribution being from APOE. Note that the asso-
and are widely used in complex
Rare variation. Before 2007, rare autosomal dominant ciation of one gene identified by GWASs, complement
trait genetics. Meta-analyses
combine the summary results
mutations in amyloid beta precursor protein (APP), component (3b/4b) receptor 1 (CR1), might result from
from each study.
presenilin 1 (PSEN1) and PSEN2 were known to cause structural variation37.
2 | ADVANCE ONLINE PUBLICATION
www.nature.com/reviews/genetics
(c) 2012 Macmillan Publishers Limited. All rights reserved

---

R E V I E W S
Figure 1 | Results pertaining to genetic architecture. a | Plot of heritability by
a 1.0
log (lifetime prevalence) for the nine psychiatric disorders considered in this Review plus
10
three complex diseases for which genetic dissection has been particularly successful
(TABLE1; Supplementary information S1 (table)). Each disorder is plotted as heritability
SCZ
by lifetime prevalence. Colour indicates qualitative success in identifying aetiological
ASD
genetic variation (with bright green meaning notably successful, dark green meaning
0.8
ADHD
some successes and red meaning minimal or no clear success to date). The bubble sizes
BIP
are proportional to the numbers of cases studied in genome-wide association studies
NIC
(GWASs; the smaller bubble indicates discovery number of cases (N ), and the larger
case
bubble indicates the total N for discovery plus replication samples). b | Allelic spectrum
case
T2DM
of schizophrenia (SCZ). The inset is a conceptual schematic from a 2008 Nature Reviews
0.6
AD
Genetics article10. The lower part of the figure depicts empirical results for SCZ.
CD
AN
The x-axis is log (allele frequency (AF)) in controls. The y-axis is the point estimate for
10
BRCA
log (genotypic relative risk (GRR). For clarity, confidence intervals are not shown. There
10
are no known Mendelian variants for SCZ (AF << 0.0001, GRR >> 50). There are no known
Heritability
ALC
common variants (AF > 0.1) with GRR > 1.5, and these can be excluded with >99%
0.4
statistical power. Nine structural variants associated with SCZ are shown as light blue
diamonds (TABLE2; 1q21.1- is the deletion and 1q21.1+ is the duplication). If AF in controls
MDD
was 0, AF was set to 0.0001. These structural variants do not have a corresponding region
in the inset. Seventeen common variants have been associated with SCZ (red circles;
TABLE3). SNPs contributing to the Psychiatric Genomics Consortium SCZ risk profile
0.2
score59 (21,171 autosomal SNPs with P < 0.1; BOX3, panel b) are shown in light blue dots
T
with a lowess smoother in dark blue. AD, Alzheimer's disease; ADHD, attention-deficit
hyperactivity disorder; ALC, alcohol dependence; AN, anorexia nervosa; ASD, autism
spectrum disorder; BIP, bipolar disorder; BRCA, breast cancer; CD, Crohn's disease; MDD,
major depressive disorder; NIC, nicotine usage (maximum cigarettes per day); SCZ,
0
schizophrenia; T2DM, type2 diabetes mellitus. The inset in panel b is adapted, with
0.001
0.01
0.1
permission, from REF.10 (c) (2008) Macmillan Publishers Ltd. All rights reserved.
log (lifetime prevalence)
10
b
50
Penetrance
High
Mendelian
Highly unusual
disease
for common
20
22q11.21
diseases
Intermediate
Low-frequency
VIPR2
15
variants with
intermediate
15q13.3
penetrance
10
Modest
Most variants
1q21.1-
16p11.2
Hard to identify
identified
genetically
by GWASs
(GRR)
NRXN1
10
Low
log
Allele
5
17q12
1q21.1+
frequency
Very rare 0.001
Rare
0.01 Uncommon 0.1 Common
3q29
2
1.5
1.2
1.0
0
0.0001
0.001
0.01
0.05
0.1
0.5
log (AF) in controls
10
Nature Reviews | Genetics
Intriguingly, pathway analyses (BOX2) of Alzheimer's metabolism had previously been proposed as Alzheimer's
disease implicate cholesterol metabolism and the innate disease risk factors, but whether these represented
immune response38. Genes attaining genome-wide sig- causation or reverse causation was unclear39. The genetic
nificance point towards immune and inflammatory findings now strongly point to reverse causation.
processes (clusterin (CLU) and CR1), lipid processing
It is unclear how the above findings relate to accu-
Lowess smoother
(APOE, CLU and ABCA7) and endocytosis (phosphati- mulation of -amyloid in Alzheimer's disease pathogen-
Locally weighted scatterplot
dylinositol-binding clathrin assembly protein (PICALM), esis, but some relationship seems likely. For example,
smoothing, which is one
technique to fit a curve to a
bridging integrator 1 (BIN1), CD2-associated protein phosphatidylinositol-binding clathrin assembly protein
scatterplot.
(CD2AP) and CD33). Altered immune function and lipid (PICALM) and other endocytic molecules can modify
NATURE REVIEWS | GENETICS
ADVANCE ONLINE PUBLICATION | 3
(c) 2012 Macmillan Publishers Limited. All rights reserved

---

R E V I E W S
-amyloid toxicity in yeast and other model systems40.
Box 1 | Study design considerations: simplex and multiplex
Although these genetic findings provide support for
Study design is a crucial component of human genetics research. The major designs
novel causal relationships that could be targeted by treat-
are case-control studies and pedigree-based studies. The most common design is
ments, the association data point to genomic regions, not
the case-control study, in which the frequency of a genetic variant in those cases
genes. Moreover, the proximal steps from genotype to
with a disorder is contrasted against the appropriate control group. Case-control
phenotype are unclear, and many of the implicated genes
designs are used in most genome-wide association studies (GWASs)156 and
are plausibly involved in multiple relevant functions (for
next-generation sequencing studies, as they are efficient and conceptually
example, CLU is involved in altered immune function
straightforward157. Case-control studies are simpler, and most biases can be
surmounted by careful study procedures, but they cannot delineate rare inherited
and lipid processing).
variation from denovo variation. Family-based designs are more complex but can be
used for association testing as well as for linkage evaluation of co-segregation of
Psychotic disorders
genotypes and phenotypes within pedigrees. They provide protection against a key
Rare variation. Unfortunately, unlike the case for
form of bias (namely, population stratification artefacts) but are less efficient given
Alzheimer's disease, no Mendelian forms of bipolar dis-
that the unit of analysis is a set of relatives; however, it is possible to identify
order (BIP) and schizophrenia (SCZ) have been identi-
mutations that arise denovo.
fied41. However, rare but potent structural variants (with
An additional decision is whether to focus on the presence or absence of other
a frequency of <0.5% and a genotypic relative risk (GRR) of
affected family members (multiplex pedigrees and simplex pedigrees, respectively).
5-20) have a role in a small proportion of cases with SCZ
Human genetic studies have classically focused on multiplex pedigrees under the
(TABLE2; Supplementary information S3 (figure)). None is
assumption that these pedigrees are enriched for causal genetic variation with
higher penetrance. A focus on multiplex pedigrees has led to the identification of
ful y penetrant, and nearly all appear to be nonspecific, as
specific mutations underlying hundreds of Mendelian disorders (including autism
risk is often increased for SCZ, autism spectrum disorder
spectrum disorder (ASD) and Alzheimer's disease). Simplex pedigrees have become
(ASD), developmental delay, intel ectual disability, epi-
popular for ASD and schizophrenia (SCZ). Simplex-based ascertainment is tailored
lepsy, somatic dysmorphism and extremes of body mass
to evaluate denovo mutations and is predicated on a model in which disorder with
and head size. Most of these structural variation regions
dramatically reduced fecundity and a proven role of denovo structural variation
are fairly large (hundreds of kilobases to hundreds of
might be explicable as a series of Mendelian disorders that can be attributed to
megabases) and general y centre on structural variation
recent high-penetrance mutations in any of a large number of genes.
hotspots42. Two rare structural variants affect single genes
However, this important choice is not simple, and it continues to be moderately
(namely, neurexin 1 (NRXN1) and vasoactive intestinal
controversial. Some investigators believe a focus on simplex pedigrees to be optimal,
peptide receptor 2 (VIPR2))43,44, offering opportunities for
and other investigators have concerns about the implications of this decision. Some of
downstream functional studies. Pathway analyses of genes
the issues are listed below.
that are intersected by rare structural variants suggest
*Correct classification as simplex or multiplex requires confident knowledge of
family history -- many people either do not know their family psychiatric histories,
enrichment for neuronal processes of plausible aetiologi-
true episodes of illness may have been kept private from other relatives, and some
cal relevance (for example, postsynaptic signalling)45-47.
affected individuals can over-call illness in their relatives (for example, an
The structural variation regions in TABLE2 probably rep-
individual with alcohol dependence labelling all relatives who drink as the same).
resent `low-hanging fruit', and more discoveries are likely
*Fecundity is a major confounder. If there are greater numbers of relatives, there
when improved technologies for structural variation
is a greater chance of multiplex classification. In addition, the presence of a
detection are applied to larger samples15.
psychiatric disorder can reduce fecundity (for example, fecundity is reduced in
A complementary approach is to evaluate structural
SCZ, and having a child with ASD can be a powerful inducement not to reproduce
variation `burden' in cases compared to controls (for
further). If fecundity had not been inhibited because of a psychiatric disorder,
example, the number of structural variants per per-
some apparently simplex families might have been revealed to be multiplex.
son)48,49. This approach tests an explicitly multigenic
*Simplex designs often require both parents. This complicates recruitment,
model whereby many rare but different genomic disrup-
increases genetic assay costs and becomes increasingly less practical for disorders
tions have an impact on disease risk. Increased struc-
with later ages of onset.
tural variation burden in SCZ cases has been reported
*Both designs have a hidden weakness in the possibility of enriching for
by multiple groups47,48,50. One report found more rare
environmental causes of illness. Many psychiatric disorders have multiple different
structural variation in SCZ cases (odds ratio = 1.15), par-
but rare environmental risk factors that are sufficient to cause a disorder. These
ticularly for large deletions (odds ratio = 3.6)48. Denovo
potent exposures are sometimes very difficult to detect or are not routinely
structural variants are also more common in SCZ cases45.
evaluated. Examples include mercury poisoning and ASD or viral meningitis and
SCZ. Contrary to its intent, simplex cases may be enriched for difficult-to-detect,
For BIP, there are reports of increased51-53 and similar
individual-specific environmental causes. Multiplex ascertainment could enrich
structural variation burden in cases versus controls54,55.
for shared environmental causes.
Denovo structural variation may be relevant in BIP (odds
Some recent data pertain to this choice. Unexpectedly, simplex and multiplex ASD
ratio = 4.8), particularly in cases with earlier ages of onset
pedigrees show fairly similar denovo mutation rates for structural variation81,82 and
(odds ratio = 6.3)53.
exonic variation84-86. It is possible that larger studies will find differences in denovo
Multiple studies are now evaluating the role of
mutation rates between simplex and multiplex families but the magnitude is likely to
denovo, rare and uncommon exonic variation in BIP
be smaller than anticipated. For SCZ, the available data are insufficient to resolve
and SCZ using resequencing or genotyping approaches.
this issue45,50. It has also been pointed out that denovo events must confer risk in
Two small-exome-sequencing studies56,57 reported rates
multiplex families, as such mutations increase the chance that an individual is
of putatively functional mutations that exceeded null
affected and increase risk in that person's offspring. Intriguingly, there are three
expectations in SCZ cases (although the rate of denovo
instances of ASD cases with denovo deletions of 16p11.2 who also had an affected
point mutations was not elevated in cases, and specific
sibling without this deletion158-160, along with similar observations for structural
variants in 1q21.1 and 17p12 (REF.159).
genes were not identified). A recent study58 conducted
high-throughput sequencing on 166 cases with SCZ and
4 | ADVANCE ONLINE PUBLICATION
www.nature.com/reviews/genetics
(c) 2012 Macmillan Publishers Limited. All rights reserved

---

R E V I E W S
Table 2 | Structural variation associated with psychiatric disorders
Structural Location
Genes Type
Disorder Frequency Frequency Odds P
Other associations
Refs
variant
(Mb)
in cases
in controls ratio value
1q21.1
chr1:
34
Deletion
SCZ
0.0018
0.0002
9.5
8 x 10-6
Developmental delay,
184
145.0-148.0
intellectual disability, micro-
Duplication SCZ
0.0013
0.0004
4.5
0.02
184
and macrocephaly, dysmorphia,
epilepsy, cataracts, cardiac
defects, possibly ASD185,
thrombocytopenia-absent
radius syndrome48,159,184-188
2p16.3
chr2:
NRXN1 Deletion
ASD
0.004
Developmental delay,
81
50.1-51.2
exons
intellectual disability, epilepsy,
Pitt-Hopkins-like syndrome 2
Deletion
SCZ
0.0018
0.0002
7.5
1 x 10-6
184
3q29
chr3:
19
Deletion
SCZ
0.0010
0.0
3.8
4 x 10-4
Developmental delay,
184
195.7-197.3
intellectual disability,
possibly ASD
7q11.23
chr7:
25
Duplication ASD
0.0011
0.003
Developmental delay,
81
72.7-74.1
intellectual disability. Deletion:
Williams-Beuren syndrome
7q36.3
chr7:
VIPR2
Duplication SCZ
0.0024
0.0001
16.4
4 x 10-5
44,
158.8-158.9
184
15q11.2
chr15:
70
Duplication ASD
0.0018
4 x 10-9
Developmental delay,
81
23.6-28.4
intellectual disability,
Prader-Willi and Angelman
syndromes188
15q13.3
chr15:
12
Duplication ADHD
0.0125
0.0061
2.1
2 x 10-4
Developmental delay,
120
30.9-33.5
intellectual disability,
Duplication ASD
0.0013
2 x 10-5
81
epilepsy188,189
Deletion
SCZ
0.0019
0.0002
12.1
7 x 10-7
184
16p13.11
chr16:
8
Duplication ADHD
0.0164
0.0009
13.9
8 x 10-4
Deletion: developmental delay,
119
15.4-16.3
epilepsy188,189
16p11.2
chr16:
29
Deletion
ASD
0.0037
5 x 10-29 Developmental delay,
81
29.5-30.2
intellectual disability, epilepsy,
macrocephaly, obesity190,191
Duplication ASD
0.0013
2 x 10-5
Developmental delay,
81
intellectual disability, epilepsy,
microcephaly, low body mass
index190,191
Duplication SCZ
0.0031
0.0003
9.5
3 x 10-8
184
17q12
chr17:
18
Deletion
ASD
0.0017
0.0
6.12
9 x 10-4
192
34.8-36.2
Deletion
SCZ
0.0006
0.0
4.49
3 x 10-4
22q11.21
chr22:
53
Deletion or ASD
0.0013
0.002
Developmental delay,
81
18.7-21.8
duplication
intellectual disability, velocard-
iofacial-DiGeorge syndrome
Deletion
SCZ
0.0031
0.0
20.3
7 x 10-13
184
Locations are US National Center for Biotechnology Information (NCBI) Build 37 and University of California, Santa Cruz (UCSC) hg19. The positions of these structural
variants are denoted in Supplementary information S3 (figure) with yellow circles. For succinctness, the citations refer to the most comprehensive study rather than to
an initial report. `Genes' refers to the number from the UCSC Known Genes data set. ADHD, attention-deficit hyperactivity disorder; ASD, autism spectrum disorder;
SCZ, schizophrenia.
on 307 controls, followed by genotyping of over 5,000 seven significant loci (TABLE3). A sign test for consist-
variants in 2,617 independent cases and 1,800 controls ency between the mega-analysis and the fol ow-up stage
-amyloid
Neuronal accumulation of this
(the cases were enriched for treatment resistance or was highly significant, implying that many of the 81 top
peptide contributes to the
strong family histories). No finding met genome-wide loci include true risk loci but that power was insufficient.
aetiology of Alzheimer's
significance. Larger studies are ongoing and will aid For BIP, the discovery phase consisted of 7,481 cases and
disease.
understanding this area in 2012-2013.
9,250 controls with a follow-up of 34 statistically inde-
Multiplex pedigrees
pendent loci in around 4,500 cases. Two loci exceeded
Family constellations
Common variation. The Psychiatric Genomics genome-wide significance (TABLE3). Similarly, a sign test
containing more than one
Consortium (PGC) recently published mega-analyses for between the discovery and fol ow-up results was highly
affected individual.
SCZ and BIP59,60. In SCZ, 9,394 cases and 12,462 controls significant, again suggesting insufficient power60.
Simplex pedigrees
were combined in a single analysis, and the top 81 stat-
In BIP, the genome-wide significant association at
Family constel ations containing
istically independent loci from that analysis were then calcium channel, voltage-dependent, L type, alpha 1C
one affected individual.
tested in over 8,000 cases. The mega-analysis identified subunit (CACNA1C) deserves specific comment, given
NATURE REVIEWS | GENETICS
ADVANCE ONLINE PUBLICATION | 5
(c) 2012 Macmillan Publishers Limited. All rights reserved

---

R E V I E W S
Table 3 | Genome-wide association study findings for psychiatric disorders
Phenotype
SNP
Location
Discovery GWAS
Largest meta-analysis
P value
Odds Nearest gene
(cases/controls)
(cases/controls)
ratio
Alzheimer's
rs3818361
chr1:207784968
2,018/5,324 (REF.34)
<19,870/39,846 (REF.35)
3.7 x 10-14
1.18
CR1
disease
rs744373
chr2:127894615
3,006/14,642 (REF.193)
<19,870/39,846 (REF.35)
2.6 x 10-14
1.17
BIN1
rs9349407
chr6:47453378
8,309/7,366 (REF.36)
18,762/29,827 (REF.36)
8.6 x 10-9
1.11
CD2AP
rs11767557
chr7:143109139
8,309/7,366 (REF.36)
18,762/35,597 (REF.36)
6.0 x 10-10
1.11
EPHA1
rs11136000
chr8:27464519
3,941/7,848 (REF.33)
8,371/26,965 (REF.193)
1.6 x 10-16
1.18
CLU
rs610932
chr11:59939307
6,688/13,251 (REF.35)
>19,000/38,000 (REF.35)
1.2 x 10-16
1.10
MS4A cluster
rs3851179
chr11:85868640
3,941/7,849 (REF.33)
8,371/26,966 (REF.193)
3.2 x 10-12
1.15
PICALM
rs3764650
chr19:1046520
5,509/11,531 (REF.35)
>17,000/34,000 (REF.35)
5.0 x 10-21
1.23
ABCA7
rs2075650
chr19:45395619
8,371/26,966 (REF.193)
1 x 10-295
2.53
APOE, TOMM40
rs3865444
chr19:51727962
8,309/7,366 (REF.36)
18,762/29,827 (REF.36)
1.6 x 10-9
1.10
CD33
Alcohol
rs1229984
chr4:100239319
REF.102
1.3 x 10-11
ADH1B
consumption
rs6943555
chr7:69806023
REF.101
4.1 x 10-9
AUTS2
rs671
chr12:112241766 REF.100
3 x 10-211
ALDH2
Bipolar
rs12576775
chr11:79077193
7,481/9,251 (REF.60)
11,974/51,793 (REF.60)
4.4 x 10-8
1.14
ODZ4
disorder
rs4765913
chr12:2419896
7,481/9,250 (REF.60)
11,974/51,792 (REF.60)
1.5 x 10-8
1.14
CACNA1C
rs1064395
chr19:19361735
682/1300 (REF.194)
8,441/35,362 (REF.194)
2.1 x 10-9
1.17
NCAN
Nicotine
rs1329650
chr10:93348120
38,181 (REF.93)
73,853 (REF.93)
5.7 x 10-10
LOC100188947
consumption
rs1051730
chr15:78894339
38,181 (REF.93)
73,853 (REF.93)
2.8 x 10-73
CHRNA3
rs3733829
chr19:41310571
38,181 (REF.93)
73,853 (REF.93)
1.0 x 10-8
EGLN2, CYP2A6
Smoking
rs3025343
chr9:136478355
41,278 (REF.93)
64,924 (REF.93)
3.6 x 10-8
1.13
DBH
cessation
Smoking
rs6265
chr11:27679916
74,035 (REF.93)
143,023 (REF.93)
1.8 x 10-8
0.94
BDNF
initiation
Schizophrenia rs1625579
chr1:98502934
9,394/12,462 (REF.59)
17,839/33,859 (REF.59)
1.6 x 10-11
1.12
MIR137
rs2312147
chr2:58222928
18,206/42,536 (REF.195)
1.9 x 10-9
1.09
VRK2
rs1344706
chr2:185778428
479/2,937 (REF.174)
18,945/38,675 (REF.196)
2.5 x 10-11
1.10
ZNF804A
rs17662626
chr2:193984621
9,394/12,463 (REF.59)
17,839/33,860 (REF.59)
4.6 x 10-8
1.20
PCGEM1
rs13211507
chr6:28257377
3,322/3,587 (REF.70)
18,206/42,536 (REF.195)
1.4 x 10-13
1.22
MHC
rs7004635
chr8:3360967
9,394/12,465 (REF.59)
17,839/33,862 (REF.59)
2.7 x 10-8
1.10
MMP16
rs10503253
chr8:4180844
9,394/12,464 (REF.59)
17,839/33,861 (REF.59)
4.1 x 10-8
1.11
CSMD1
rs16887244
chr8:38031345
3,750/6,468 (REF.68)
8,133/11,007 (REF.68)
1.3 x 10-10
1.19
LSM1
rs7914558
chr10:104775908 9,394/12,466 (REF.59)
17,839/33,863 (REF.59)
1.8 x 10-9
1.10
CNNM2
rs11191580
chr10:104906211 9,394/12,467 (REF.59)
17,839/33,864 (REF.59)
1.1 x 10-8
1.15
NT5C2
rs11819869
chr11:46560680
1,169/3,714 (REF.197)
3,738/7,802 (REF.197)
3.9 x 10-9
1.25
AMBRA1
rs12807809
chr11:124606285
18,206/42,536 (REF.195)
2.8 x 10-9
1.12
NRGN
rs12966547
chr18:52752017
9,394/12,468 (REF.59)
17,839/33,865 (REF.59)
2.6 x 10-10
1.09
CCDC68
rs9960767
chr18:53155002
18,206/42,537 (REF.195)
4.2 x 10-9
1.20
TCF4
its mechanistic implications. Indeed, multiple voltage calcium sensor62. Therefore, a detailed investigation of
gated calcium channel subunits were among the top 34 brain calcium biology is warranted for both BIP and SCZ.
loci fol owed up in the BIP GWAS. Calcium channels are
For SCZ, the strongest association is in the extended
Genotypic relative risk
a treatment for BIP and regulate neuronal excitability and major histocompatability complex region (MHC region;
(GRR). A measure of the effect
multiple brain functions, including long-term potentia- chr6:27-33 Mb). The evidence for association is compel-
size of a genetic variant ranging
tion and synaptic plasticity. Combined analysis of the ling, but high gene density and exceptional y high linkage
from zero to infinity. A GRR of 1
PGC BIP and SCZ samples strengthened the association disequilibrium complicate the identification of specific
means no change in risk,
GRR < 1 is protective and
in the CACNA1C region. Further, results from SGENE+61 sequence variation. Although it is tempting to propose
GRR > 1 is predisposing.
implicate neurogranin (NRGN), which may act as a that the association supports long-standing hypotheses
6 | ADVANCE ONLINE PUBLICATION
www.nature.com/reviews/genetics
(c) 2012 Macmillan Publishers Limited. All rights reserved

---

R E V I E W S
Table 3 (cont.) | Genome-wide association study findings for psychiatric disorders
Phenotype
SNP
Location
Discovery GWAS
Largest meta-analysis
P value
Odds Nearest gene
(cases/controls)
(cases/controls)
ratio
Schizophrenia rs1344706
chr2:185778428
479/2,937 (REF.174)
21,274/38,675 (REF.196)
4.1 x 10-13
1.11
ZNF804A
and bipolar
disorder
rs2239547
chr3:52855229
9,394/12,471 (REF.59)
16,374/14,046 (REF.59)
7.8 x 10-9
1.12
ITIH3-ITIH4
rs10994359
chr10:62222107
9,394/12,470 (REF.59)
16,374/14,045 (REF.59)
2.4 x 10-8
1.22
ANK3
rs4765905
chr12:2349584
9,394/12,469 (REF.59)
16,374/14,044
7.0 x 10-9
1.11
CACNA1C
This table focuses on results achieving genome-wide significance in large samples. We use a significance threshold of 5 x 10-8 (REF.198). Most associations that
achieve this level of significance are secure, but some may ultimately prove not to be. Included are SNPs with P < 5 x 10-8 that were evaluated in samples of a
minimum of around 10,000 cases and 10,000 controls. Discovery sample sizes reflect the primary samples for which full genome-wide association studies (GWASs)
were conducted. In most cases, discovery P values were > 5 x 10-8 but met a threshold (typically 1 x 10-5) for inclusion in replication efforts. In some instances,
simultaneous publications based on overlapping samples were considered to be `discovery' studies. Where this occurred, providing samples of roughly equivalent
sizes, the most significant primary GWAS findings are given, otherwise the largest discovery samples are favoured. For many studies, it was not possible to extract
the exact sample size used for each locus, so the sample sizes above are approximate. P values and odds ratios are from the meta-analysis with the largest sample
sizes. If two meta-analyses based on overlapping samples reported similar results, the `discovery' study is cited. The genes that are nearest to each locus are
provided, but aetiological variants are generally unknown, and it remains likely that some of the associations do not alter the function of the designated genes (for
example, ITIH3-ITIH4, in which multiple correlated SNPs span many genes, and TCF4-CCDC68, in which statistically independent associations occur in TCF4 and
closer to CCDC68). In the TCF4-CCDC68 example, it may be that both associations point to the same functional element, but it is also possible that independent
aetiological variants occur in adjacent genes. For the schizophrenia loci attributed to REF.195, no discovery sample size is listed because the initial P values were
modest, and as the authors conducted multiple follow-up analyses, there is no obvious discovery sample. For the major histocompatibility complex (MHC), the
International Schizophrenia Consortium70 is designated a discovery study, as it was the only primary GWAS for which genome-wide significance at the MHC was
obtained. REF. 195 is cited for the meta-analysis at the MHC, as it reported the most significant MHC association195. The most significant SNP at the MHC across
the two studies is not identical, and the one listed is taken from REF.195. Multiple statistically independent SNPs have been reported at the MHC59,186. We note that
genome-wide significance had been reported in bipolar disorder for ankyrin 3 (ANK3) (REF.199) but not in a larger mega-analysis that included the same samples60.
Others have reported genome-wide significance for composite phenotype studies of ITIH3-ITIH4 (REF.200) (but see REF.201) and calcium channel,
voltage-dependent, L type, alpha 1C subunit (CACNA1C)202 but in samples smaller than required for inclusion in the above table. For alcohol consumption, the
rs671 association was in East Asian samples. The alcohol dehydrogenase 1B (class I), beta polypeptide (ADH1B) locus was also associated with alcohol dependence.
Nicotine consumption is measured in terms of maximal number of cigarettes smoked per day. Smoking initiation refers to ever versus never began smoking.
Smoking cessation is whether regular smokers had quit at the time of interview.
concerning roles in SCZ for intra-uterine infection, auto- Autism spectrum disorder
immunity or even synaptic pruning (in which MHC Rare variation. For ASD, there is a notably strong prima
genes have a role), this lack of precision renders such facie case for there being a cardinal role for rare vari-
propositions speculative.
ation. Karyotyping studies suggest that on the order of
A novel association for SCZ is in Ensembl gene 5% of ASD cases have one of a large number of rare but
RP11-490G2.1, which encodes the primary transcript fairly gross chromosomal abnormalities14,79. In addition,
for the microRNA miR-137 (MIR137)63. Supporting the ASD has been noted as a co-morbid feature of >100
hypothesis that this association implicates MIR137, pre- single-gene Mendelian medical genetic syndromes80,
dicted targets of miR-137 were significantly enriched for although the penetrance and confidence of the clinical
smal er GWAS P values (P < 0.01), and four of the genes associations are variable. Indeed, ASD mutations with a
Odds ratio
that achieved genome-wide significance contain verified high penetrance are exceptional (that is, Rett's syndrome
Similar to genotypic relative
miR-137-binding sites64. miR-137 is a key regulator of mutations in methyl-CpG-binding protein 2 (MECP2)
risk: a measure of the change
neuronal development with roles in neurogenesis and and cyclin-dependent kinase-like 5 (CDKL5)), and
in risk associated with a
maturation65,66 and is highly expressed at synapses in the Mendelian diseases that are enriched for ASD have far
genetic variant.
cortex and hippocampus67. Future studies of networks less than complete penetrance (for example, fragile X
Mega-analyses
regulated by miR-137 offer the possibility of insights into syndrome and tuberous sclerosis)79.
These are methods (that
SCZ pathophysiology.
Analysis of structural variation has been a major focus
are less widely used than
GWASs of BIP and SCZ have been predominantly in ASD research (TABLE2; Supplementary information S3
meta-analyses) for
summarizing and combining
based on subjects of European ancestry, but there are (figure)). Implicated loci to date are generally rare and
results across multiple studies.
increasing reports from other world ancestries68,69. potent risk factors but are incompletely penetrant
Mega-analysis combines
Although those findings do not yet provide additional and are not specific to ASD. As these large events have an
individual-level genotype and
pathophysiological insights, it is worth noting that impact on the dosages of many genes, biological insight
phenotype data from all
a chromosome 8 locus found in an East Asian sam- has been slow to emerge; however, pathway analyses of
subjects in each study.
ple68 has support in the PGC data set, suggesting that genes within structural variants do implicate neuronal
Major histocompatibility
planned mega-analyses across world populations will be processes of aetiological relevance45-47. Large structural
complex region
informative.
variants are present in 5-10% of ASD cases, and the
(MHC). A region of
Some of the most intriguing findings for SCZ and BIP number of ASD structural variants could total 130-234
approximately 3 Mb on human
chromosome 6p22.1 that is
are from large sets of genetic markers70 (BOX3). There are (REF.81). There is also consistent evidence for increased
exceptionally complex and has
now replicated data that vulnerability to SCZ is influ- structural variation burden in ASD49,81-83. For example,
considerable importance to
enced by common genetic variation in hundreds of dif- 5.8% of ASD probands had 1 rare denovo structural
disease. It contains genes
ferent loci, and this vulnerability partial y overlaps that variant versus 1.7% of their unaffected siblings (odds
encoding cell surface molecules
for BIP70. Indeed, the large-scale impact of large numbers ratio = 3.5), and this difference was more pronounced
that are important to immunity
and disease susceptibility and
of common variants may be a general feature of human for structural variants that intersected genes81. The
many other functions.
complex traits71-78
16p11.2 structural variant that is associated with ASD
NATURE REVIEWS | GENETICS
ADVANCE ONLINE PUBLICATION | 7
(c) 2012 Macmillan Publishers Limited. All rights reserved

---

R E V I E W S
and SCZ has been termed a `mirror image' structural
Box 2 | Pathway analysis
variant, as the deletion and duplication are associated
Pathway analysis is based on the assumption that risk variants for a disease will
with increased and reduced head and body size. However,
converge on sets of genes with functions that are more closely related to each other
it is difficult to understand the clinical features of ASD
than to random sets of genes. For example, dominant forms of Alzheimer's disease are
and SCZ as mirror images, and more importantly ASD is
caused by mutations in amyloid beta precursor protein (APP), presenilin 1 (PSEN1) and
associated with both 16p11.2 deletions and duplications.
PSEN2; the PSEN genes encode protein components of -secretase, a protease that
ASD is the first psychiatric disorder for which exome
cleaves APP. The availability of genome-wide association study (GWAS) and structural
sequencing using substantial numbers of samples has
variation results for many psychiatric disorders, along with increasing amounts of
sequence data, have generated interest in using analytic methods for exploiting
been published. Three recent papers describe the results
nonrandom functional relationships between genes containing risk variants. Many
from exome sequencing of ~600 trios and identify roles
approaches have been developed (for example, ALIGATOR161, INRICH162, DAPPLE163 and
for denovo exonic mutations in sodium channel protein
GRAIL164) and reviewed in detail elsewhere147,165,166. Although the algorithms differ, the
type 2 subunit alpha (SCN2A), katanin p60 subunit
principle behind these methods is to evaluate whether a given set of genomic regions
A-like 2 (KATNAL2) and chromodomain helicase
(that is, a broadly inclusive definition of `pathway') is enriched for genetic variants that
DNA-binding protein 8 (CHD8) in the pathogenesis of
show some relationship with disease compared to a null expectation.
ASD84-86. Intriguingly, al three studies noted an increased
There are important subsidiary considerations. The first is the definition of a
rate of denovo exonic mutations in older parents (with
`pathway'. Standard pathways consist of sets of genes found in the Gene Ontology167,
the mutations general y being of paternal origin)84-86, and
the Kyoto Encyclopedia of Genes and Genomes168 or PANTHER databases169. Other
pathway analyses reported in two of the studies found
pathway gene sets are manually curated by experts in a particular area (for example,
genes that are known to make proteins that function at the synapse)170,171. In addition,
that genes containing denovo exonic variation were more
a `pathway' can consist of genomic regions that are selected for a particular property,
closely connected in reference to protein-protein inter-
such as a high degree of conservation172 or expression quantitative trait locus (eQTL)
action databases84,85. Additional sequencing studies are
associations173. Finally, other pathways consist of genes that are known to be connected
in progress.
by experimental data (for example, by protein-protein interaction screens, microRNA
However, a central finding from these papers was that
target sites or gene expression modules).
only a minority of cases had a denovo putatively func-
It is advantageous that these pathway data sets are defined independently of genetic
tional variant, suggesting that this class of genetic variation
studies of psychiatric disorders, but they do have limitations and can contain errors of
is unlikely fully to explain the clinical entity of ASD.
omission and commission. Standard gene sets can have highly overlapping pathways
Indeed, estimates from denovo exonic mutations (which
that complicate some analyses, and the pathway content can have variable quality.
are similar to those from structural variation data) sug-
Expert-curated pathways can be more specific but can be vulnerable to post hoc bias
gest that ASD is highly polygenic (estimates ranged from
(that is, including genes in a pathway on the basis of results from genetic studies).
Pathways based on empirical approaches depend on the quality and completeness of
400-1,000 genes)85,86. Importantly, a hypothetical model
the primary data (for example, existing protein-protein interaction databases cover the
of ASD that it is caused by rare but ful y penetrant muta-
interaction space partially).
tions in 100 different genes could be confidently rejected84.
A second question concerns what is required before a member of a pathway is
accepted as having some relationship with disease. For common variation, the analysis
Common variation. Evaluation of rare structural varia-
might be restricted to genes within recombination regions containing SNPs that have
tion and exonic variation in ASD is particularly advanced.
genome-wide significance, an approach that was used successfully to implicate broad
By contrast, evaluation of common variation is far more
biological pathways that are relevant to height76. However, much of the interest in
limited (FIG.1a), and the published GWASs for ASD are
pathway analysis involves exploiting much weaker associations under the assumption
small by current standards87-90. It is currently not possible
that these associations more reflect true associations in the context of limited power
to discern or dismiss a role for common genetic varia-
(signal) rather than chance (noise). If so, those weakly associated SNPs may also be
nonrandom with respect to gene sets (BOX3). The threshold at which SNPs or genes are
tion in risk for ASD. In our opinion, GWASs with larger
chosen is arbitrary, and the signal-to-noise ratio for a given arbitrary threshold can vary
samples are needed for ASD, given that detailed studies of
substantially with sample size and genetic architecture. For rare variants in complex
rare variation currently explain a fraction of risk and that
diseases, in light of recent empirical results, pathway analysis will by necessity be based
common variation plays a clear part in other psychiatric
on sets of genes for which the involvement in disease is unclear (for example, genes
disorders. Indeed, there were few confident findings for
with a single observed denovo exonic deleterious mutation)84,85.
GWASs of SCZ when the sample sizes were similar to
A third consideration concerns the null expectation to which the observed pathways
those that are now available for ASD. Additional support
are compared. Early structural variant pathway analysis did not fully account for
for our recommendation for more GWASs is provided
important biases, such as that large genes are more likely to be intersected by copy
by Voineagu etal.91, who identified a gene expression
number variants by chance and that some functional pathways -- often related to
module that had attenuated expression in post-mortem
brain development -- are enriched for large genes164. Early GWAS pathway analysis
brain samples of individuals with ASD and that also had
sometimes did not fully allow for the variable numbers of SNPs per gene and their
degree of linkage disequilibrium, both of which have an impact on the probability of a
enrichment for GWAS signals91.
high-ranking association161. Thus, it is necessary to be cautious about the use of
pathway-based approaches. In psychiatric disorders, some results give cause for
Alcohol and nicotine dependence
optimism38,45,82.
Alcohol dependence (ALC) and nicotine depend-
Finally, in pathway analyses, the unit of inference is the pathway. Tempting although
ence (NIC) are complex conditions to study, given the
it may be, it is generally inappropriate to make strong inferences about specific
requirement for ingestion of a psychoactive substance
variants or genes on the basis of their membership of pathways that attain some
and cohort effects due to temporal and geographic vari-
level of significance. It may be possible to do so if the variants or genes are
ation in the availability of ethanol and nicotine. Many
subsequently evaluated in data sets independently of those from which the importance
investigators focus on ALC and NIC, which are clinical y
of the pathways are derived, using a statistical framework that adequately deals with
salient but multi-component syndromes92. As a part of
multiple testing.
the Tobacco and Genetics Consortium93, we determined
8 | ADVANCE ONLINE PUBLICATION
www.nature.com/reviews/genetics
(c) 2012 Macmillan Publishers Limited. All rights reserved

---

R E V I E W S
Box 3 | Common variant risk profile
For schizophrenia (SCZ) and bipolar disorder (BIP), sign
a
P = 2 x 10-28
tests comparing the consistency of association tests from
discovery genome-wide association studies (GWASs) and
P < 0.1
0.03
T
replication samples for sets of top signals are usually
P < 0.2
T
highly significant even if most loci do not meet
P < 0.3
T
genome-wide significance59,60,174. This implies that sample
P < 0.4
T
sizes are insufficient and that additional loci can be
P < 0.5
T
discovered in larger samples. Tests of the existence of
5 x 10-11
large numbers of true but weakly associated variants have
been conducted for SCZ, BIP and many other biomedical
disorders.
1 x 10-12
In light of theoretical work by Visscher and colleagues71,
0.02
)
one study used GWAS results as a discovery set (after
2 R
removing correlated SNPs), and subjects in 11 independent
test GWAS data sets were assigned risk profile scores (that
is, the number of SCZ risk alleles weighted by their effect
7 x 10-9
sizes in the discovery set). The mean risk profile scores for
cases were compared to the mean scores for controls in
these independent data sets. Panel a of the figure shows risk
profile scores for extremely relaxed P value thresholds
V
ariance explained (
(P < 0.1, 0.2, 0.3, 0.4 and 0.5, light to dark green bars). Risk
0.01
T
profile scores were derived (after linkage-disequilibrium-
based SNP pruning) from a discovery SCZ sample and then
applied to three independent SCZ samples174,175, two BIP
samples176,177 and six non-psychiatric diseases177. In three
independent GWASs, SCZ cases had significantly higher
0.008
risk profile scores than controls. Remarkably, the same set
of markers also discriminated BIP cases from controls,
0.71 0.05 0.30 0.65 0.23 0.06
indicating substantial genetic contributions between SCZ
and BIP. As an important test of specificity, the SCZ risk
0
profile was not predictive of case status for any of six
CAD
CD
HT
RA
T1D
T2D
non-psychiatric diseases177. A recent paper evaluated risk
MGS-EA MGS-AA
STEP-BD WTCCC
O'Donovan
profile scores in a trio sample and could exclude population
stratification as an explanation178.
SCZ
BIP
Non-psychiatric (WTCCC)
The proportion of variance explained by the risk profile
score increased with relaxation of the significance
b 0.07
thresholds. This suggests that the discovery sample was
insufficiently large to identify many true risk loci at even
nominal levels of significance: adding more SNPs
contributed more genetic signal than noise. This is partly a
0.06
1e-60 1e-60 1e-60 1e-60
1e-50
feature of sample size. Panel b of the figure shows a similar
analysis based on a larger discovery sample, in which the
1e-50
proportion of variance explained is approximately double59,
1e-50
and instead of increasing with P the proportion of variance
0.05
T
reaches a plateau. If the sample size were truly adequate,
the first P bin would explain the greatest amount of
T
variance, and relaxing P would decrease the proportion
T
0.04
of variation explained.
Finally, estimates from two different methods indicated
2
1e-16
R
that the risk profile component for SCZ contributes
between one-quarter and one-third of the overall variance
0.03
in liability to SCZ70,131, a substantial fraction of the 65-81%
heritability of SCZ179,180. These estimates suggest that
`missing heritability' is merely hidden and imperfectly
assayed by current genotyping technologies. CAD, coronary
0.02
artery disease; CD, Crohn's disease; HT, hypertension;
1e-16
MGS-AA, Molecular Genetics of Schizophrenia -- African
American; MGS-EA, Molecular Genetics of Schizophrenia --
0.01
1e-8
European American; RA, rheumatoid arthritis; STEP-BD,
Systematic Treatment Enhancement Program for Bipolar
Disorder;T1D,type1diabetes;T2D,type2diabetes;
WTCCC, Wellcome Trust Case Control Consortium.
0
Panel a of this figure is modified, with permission, from
REF. 70 (c) (2009) Macmillan Publishers Ltd. All rights reserved.
0.0001
0.001
0.01
0.05
0.1
0.2
0.3
0.4
0.5
1
Data in panel b are taken from REF. 59.
PT
Nature Reviews | Genetics
NATURE REVIEWS | GENETICS
ADVANCE ONLINE PUBLICATION | 9
(c) 2012 Macmillan Publishers Limited. All rights reserved

---

R E V I E W S
that the components of the Fagerstrom test for nicotine in progress (for example, by the Wellcome Trust Case-
dependence (which is a measure of NIC) had heritabili- Control Consortium for anorexia nervosa). Given low
ties ranging from fairly high to near zero with impor- power, no conclusions about common variation can be
tant common environmental effects. Other investigators made. In ADHD, increased structural variation burden
evaluated self-reported lifetime maximum use of ethanol has been reported (odds ratio = 2.1)119,120, an effect that is
(measured in grams per day) or nicotine (measured in higher in ADHD cases with intel ectual disability (odds
number of cigarettes per day), and such continuous meas- ratio = 5.7)119. Pathway analysis in ADHD found associa-
ures of consumption are often available for secondary tion signals enriched in the same Gene Ontology catego-
analysis of samples studied for other diseases.
ries also overrepresented for large structural variants121.
For ALC, the published GWASs are small, and no The weak signals in ADHD GWASs are not randomly
large-scale meta-analysis has been conducted94-98. In our distributed but index the same pathophysiological path-
opinion, there are clear needs for a high-quality meta- ways as rare structural variants. Thus, it appears that the
analysis and for increasing the number of samples with reason no common variants are yet to be confidently
GWAS data -- particularly given that risk profile analysis implicated in ADHD by GWASs is a lack of power and
(BOX3) suggested that larger samples would yield more not a lack of variants to be found.
associations98. For alcohol consumption, GWASs in East
Asian samples confirmed the role of aldehyde dehydro- What is the emerging picture?
genase 2 (ALDH2)99,100, and autism susceptibility candi- Knowledge of psychiatric genetics is vastly greater than
date 2 (AUTS2) was implicated in alcohol consumption it was 5 years ago. Specifically, there are now multiple
in European subjects101. Using a candidate gene approach, high-confidence structural variants (TABLE2), rare exonic
the association of alcohol dehydrogenase 1B (ADH1B) variants (currently only for ASD, Alzheimer's disease and
with ALC and alcohol consumption was extended to ALC) and an increasing number of robustly significant
European ancestry subjects102.
and replicated common variants (TABLE3). The data sup-
For NIC, a field-wide m