Available online at www.sciencedirect.

com

ScienceDirect

Hidden hypotheses in ‘hypothesis-free’ genome-wide epigenetic

associations
Edward D Barker1, Susanna Roberts1 and Esther Walton2

The recent interest in epigenetics within mental health choices [5] can be observed across the life span, and that
research, from a developmental perspective, stems from the these long-term epigenetic modifications are associated
potential of DNA methylation to index both exposure to with risk for a range of health outcomes [6]. These studies
adversity and vulnerability for mental health problems. have generally focused on DNA methylation (DNAm) for
Genome-wide technology has facilitated epigenome-wide two reasons: it is currently the best understood epigenetic
association studies (EWAS), permitting ‘hypothesis-free’ mechanism and array-based technologies are readily
examinations in relation to adversity and/or mental health available, which provides coverage of hundreds of thou-
problems. In EWAS, rather than focusing on a priori established sands of methylation sites across the genome [7]. This
candidate genes, the genome is screened for DNA methylation, combination of basic science and genome-wide technol-
thereby enabling a more comprehensive representation of ogy has facilitated numerous epigenome-wide association
variation associated with complex disease. Despite their studies (EWAS), permitting ‘hypothesis-free’ examina-
‘hypothesis-free’ label, however, results of EWAS are in fact tions in relation to adversity and/or mental health
conditional on several a priori hypotheses, dictated by the problems.
design of EWAS platforms as well as assumptions regarding
the relevance of the biological tissue for mental health The logic underlying EWAS is comparable to genome-
phenotypes. In this short report, we review three hidden wide association studies (GWAS [8]). Rather than focus-
hypotheses — and provide recommendations — that ing on DNAm in proximity to candidate genes, the
combined will be useful in designing and interpreting EWAS genome is screened for DNAm, thus enabling a more
projects. comprehensive representation of variation associated
with complex disease. As with GWAS (e.g. [9,10]),
Addresses
1
despite their ‘hypothesis-free’ label, results of EWAS
Department of Psychology, Institute of Psychiatry, Psychology and are in fact conditional on several a priori hypotheses,
Neuroscience, King’s College London, UK
2
Medical Research Council Integrative Epidemiology Unit and Bristol
dictated by the design of EWAS platforms as well as
Medical School, Population Health Sciences, University of Bristol, UK assumptions regarding the relevance of the biological
tissue for the mental health phenotypes under investiga-
Corresponding author: Barker, Edward D ([email protected]) tion. In this short report, we review three hidden hypoth-
eses (see Figure 1) — and provide recommendations —
Current Opinion in Psychology 2019, 27:13–17 that combined will be useful in designing and interpret-
ing EWAS projects.
This review comes from a themed issue on Genetics
Edited by Brian Boutwell and Michael A White

Hidden hypothesis 1: EWAS coverage is

sufficient for complex psychiatric problems
https://doi.org/10.1016/j.copsyc.2018.07.009 Array-based platforms have become widespread in psy-
chology research, largely due to their ease of use, rela-
2352-250X/ã 2018 The Authors. Published by Elsevier Ltd. This is an
open access article under the CC BY-NC-ND license (http://creative- tively high through-put, and well standardised and vali-
commons.org/licenses/by-nc-nd/4.0/). dated pipelines for processing, quality control, and
analysis techniques. In particular, the Illumina 450k
and EPIC arrays feature 480 000–850 000 probes targeting
nearly 99% of RefSeq genes, as well as a range of other
Understanding the biological mechanisms by which early
genomic categories, such as CpG islands, shores and
psychosocial adversity associates with long-term mental
shelves, miRNA promoters and enhancers, where DNAm
health problems may have the potential to facilitate the
can be influenced by and/or impact transcription in distal
development of effective screening, intervention strate-
genomic regions [11]. Compared with the Ilumina 450k,
gies and health policy decisions [1]. Recent research has
the newer Illumina EPIC 850k array provides much
focused on the degree to which adversity disrupt gene
greater coverage of ENCODE and FANTOM5 enhan-
regulation through epigenetic processes, thereby provid-
cers [12], and shows higher genetic influence underly-
ing a mechanism by which the environment can have
ing DNAm probes [13]. Nevertheless, these microarrays
lasting effects on measurable mental health phenotypes
are limited in the number of sites they can assess, and thus
[2]. High profile studies suggest that epigenetic changes
lack true genome-wide measurements [14].
associated with early adversities [3,4] and even lifestyle

www.sciencedirect.com Current Opinion in Psychology 2019, 27:13–17

14 Genetics

Figure 1

3
2
Epithelial Blood

M covered

M relevant

M covered / relevant

Current Opinion in Psychology

Hidden hypotheses in epigenome-wide approaches. Note: (1) = Hypothesis 1: EWAS coverage is sufficient for complex psychiatric problems;
(2) = Hypothesis 2: peripheral tissue is meaningful for mental health problem(s); and (3) = Hypothesis 3: biology can be meaningful to phenotype of
interest.

Furthermore, during the design process of the 450k and disorder or depression, however, the brain is often the
EPIC arrays, CpG sites were chosen as potentially bio- main tissue of interest when it comes to mechanistic
logically informative based on consultation with a consor- interpretations of results [16]. To this end, research
tium of DNA methylation experts [15]. Whilst the cover- suggests that the correspondence of methylation profiles
age of genes and CpG islands on these microarrays are from blood and saliva to the brain is in fact quite limited,
comprehensive, it does not represent a complete picture but can be higher with cross-tissue genetic influence
of methylated cytosines across the genome. Selection [13,17]. This presents a critical disadvantage if the inves-
was, in part, based on data from a number of phenotypes tigator would like to use the peripheral tissue as a surro-
(some medical in nature such as cancer), and thus is not gate of the central nervous system (CNS; the brain).
specifically targeted to brain-based, stress-related com-
plex mental health phenotypes. This is an important One promising avenue is to establish DNAm as a bio-
point: if a sizeable proportion of the CpG sites tested marker for mental illness. A biomarker does not have to
are not relevant to the phenotype of interest, the likeli- be mechanistic (i.e. CNS surrogate). Indeed, blood-based
hood of detecting relevant results is reduced. biomarkers have been used for diagnostics, predictive
risk, disease monitoring and/or treatment response in
Hidden hypothesis 2: peripheral tissue is cancer, cardiovascular and infectious disease [18,19].
meaningful for mental health problem(s) However, even within a biomarker framework, the
The second hidden hypothesis relates to the tissue that is assumption is often that distinct peripheral tissues are
used to quantify DNAm. The majority of mental health interchangeable and equally suited for biomarker detec-
research is based on DNAm profiles obtained from tion, when in fact it is highly probable that peripheral
peripheral tissues from living persons, such as blood tissues themselves correspond differently to environmen-
and saliva. When investigating outcomes such as conduct tal adversity and/or disease state [14]. For instance,

Current Opinion in Psychology 2019, 27:13–17 www.sciencedirect.com

 Hidden hypotheses Barker, Roberts and Walton 15

biomarkers for mental health traits (e.g. depression) may The second recommendation for optimising the use of
be more detected in blood than saliva, as blood is more EWAS CpGs is to target those probes with underlying
central to inflammatory processes related to stress and genetic influence — methylation quantitative trait loci
disease [16]. (mQTLs). This approach may have the advantage that
cross-tissue concordance (e.g. blood, saliva, post-mortem
brain) appears higher for CpGs that show cross-tissue
Hidden hypothesis 3: biological relevance for
genetic influence [13]. Another advantage of mQTLs is
the phenotype of interest
that CpGs under considerable genetic influence are less
The last hypothesis relates to the assumption that biology
affected by confounds [11,13,25]. However, while
can be informative to the phenotype itself. Focal pheno-
mQTLs are a worthwhile approach, it is a relatively
types (e.g. oppositional defiant disorder, anxiety) in men-
new area and at present, there is a small proportion of
tal health research are often complex and multiply deter-
methylation sites with consistently reported mQTLs
mined [20]. The lack of established robust biomarkers for
[11,13]. Furthermore, large-scale and detailed informa-
mental health problems (e.g. [19]) may suggest that some
tion on tissue-specific mQTLs is still sparse.
of these traits might not strongly associate with detectable
biological processes. Furthermore, effect size associations
Recommendation for hidden hypothesis 2:
in EWASes are often very small suggesting that — while
peripheral tissue and phenotype
significant — distinguishing the importance of DNAm in
One strategy to maximise the interpretability of EWAS
the aetiology of the mental health phenotype may prove
projects is to examine DNAm as a biomarker for mental
difficult [21]. Perhaps unsurprisingly then, most EWAS
health problems that have mechanistic underpinning in
in mental health include some form of gene ontology
tissues other than the brain, such as blood. A wide-range
analysis, which queries the role of larger biological sys-
of psychiatric disorders have been associated with
tems based on existing databases [22]. These analyses
immune function as measured by peripheral inflamma-
result in general statements such as ‘neurodevelopment’
tion [26]. Furthermore, there is good evidence from
or the ‘immune system’ being involved in the aetiology of
animal studies, and increasing evidence in humans, that
a given phenotype. Whether these broad categories play
peripheral inflammatory markers can affect brain areas
indeed a substantial role in the aetiology of the mental
implicated in certain psychiatric disorders [27]. Conse-
health problem is often hard to determine given the post
quently, adversity-related immune processes and DNAm
hoc nature of the interpretation. Relatedly, many
may be well measured in blood samples (see [28]). For
EWASes have tried to infer downstream effects of
biomarkers to be useful, they must be cost effective,
observed variation in DNAm such as differences in gene
drawn from accessible tissue and predictive of future risk
expression. Many of these studies find very little in terms
[29]. Biomarkers for brain-based disorders (e.g. depres-
of functional relationships, but a small number do report
sion) have proven more difficult to establish [19]. Liu et al.
downstream biological associations (e.g. [21]). In gen-
[30] performed an EWAS on blood tissues across 13 pop-
eral, it has proven difficult to pinpoint EWAS-related
ulation-based cohorts and reported that a composite bio-
biological relevance of observed DNAm changes, even
marker (consisting of 144 CpGs) discriminated drinkers
if they are in genes which seem ‘plausible’ based on
from non-drinkers. It was thus suggested that a blood-
reported functionality and previous literature.
based DNAm diagnostic test could be developed. It is
important to note, however, that in addition to methodo-
Recommendation for hidden hypothesis 1: logical considerations [31], the Liu et al. study was cross-
EWAS coverage sectional, thus it may prove difficult to use this specific
An alternative to using arrays with limited coverage is to biomarker as a predictor of future alcohol use, as the
use next-generation sequencing-based approaches to variation in DNAm may be the result of chronic drinking
interrogate the whole methylome [21]. However, these (i.e. reverse causality [32]). Importantly, large-scale meta-
methodologies are high in cost and time intensive. analyses based on new and growing consortia (e.g. PACE
Despite the limitations described above, pragmatic and [33]) are beginning to report consistent epigenetic
strategic study design can maximise utility and interpre- effects on traits such as schizophrenia or smoking behav-
tation of results of the Illumina 450k and EPIC arrays. For iour (e.g. [34,35]) which suggests that we may begin to be
example, for researchers interested in targeting CpGs able to utilise this information to further optimize DNAm
likely to associate with ‘brain-based’ mental illnesses, biomarker approaches.
an a priori set of CpGs (e.g. a ‘systems approach’) could
be isolated from the array data, which could still span Recommendation for hidden hypothesis 3:
thousands of loci. The suggestion is to prioritize CpGs phenotype and biology
within biological systems that are known to associate with Several suggestions have been put forward to address the
variation in post-mortem brain samples [23] or even complex nature of the biology that may underlie mental
structural or functioning brain imaging [24] if this is of health problems. Most notably, the Research Domain
primary interest to the investigator. Criteria (RDoC) initiative has proposed alternative

www.sciencedirect.com Current Opinion in Psychology 2019, 27:13–17

16 Genetics

approaches to study mental illness by integrating many 2. Barker ED, Walton E, Cecil CAM: Annual research review: DNA
 methylation as a mediator in the association between risk
different levels of information including genetics, neu- exposure and child and adolescent psychopathology. J Child
rocircuits and behaviour [36]. Methylation-based research Psychol Psychiatry 2018, 59:303-322 http://dx.doi.org/10.1111/
jcpp.12782.
can be integrated into an RDoC perspective. Here, Very good review on the degree to which DNA methylation links risk
researchers could employ a two-stage analysis, first inves- exposure and psychopathology.
tigating epigenetic effects on intermediate dimensions of 3. Klengel T, Mehta D, Anacker C, Rex-Haffner M, Pruessner JC,
mental health and then, using the results as biomarkers to Pariante CM, Pace TW, Mercer KB, Mayberg HS, Bradley B et al.:
Allele-specific FKBP5 DNA demethylation mediates gene–
query the more complex phenotypes. For, example, if childhood trauma interactions. Nat Neurosci 2013, 16:33.
externalising difficulties (e.g. ADHD, aggression) are the 4. Rijlaarsdam J, Cecil CA, Walton E, Mesirow MS, Relton CL,
focal phenotype, rather than performing an EWAS Gaunt TR, McArdle W, Barker ED: Prenatal unhealthy diet,
directly on the disorder(s), the researchers could instead, insulin-like growth factor 2 gene (IGF2) methylation, and
attention deficit hyperactivity disorder symptoms in youth
as the first step, perform an EWAS on brain imaging with early-onset conduct problems. J Child Psychol Psychiatry
endophenotypes of the externalising phenotype (e.g. 2017, 58:19-27.
[24]). In the second step, the results of the EWAS could 5. Richmond RC, Simpkin AJ, Woodward G, Gaunt TR, Lyttleton O,
be used to create poly-epigenetic genetic biomarker score  McArdle WL, Ring SM, Smith AD, Timpson NJ, Tilling K: Prenatal
exposure to maternal smoking and offspring DNA methylation
(e.g. [28]) to be (potentially) associated with the disor- across the lifecourse: findings from the Avon Longitudinal
der. This type of two stage of EWAS may examine the Study of Parents and Children (ALSPAC). Hum Mol Genet 2015,
24:2201-2217.
epigenetic changes associated with antecedents of diag- Longitudinal study describing the transient, but also partially persistent
nosable mental health conditions, which would be could effects of maternal smoking on offpsring DNA methylation.
be more useful as a risk biomarker than a biomarker of the 6. Tobi EW, Slieker RC, Luijk R, Dekkers KF, Stein AD, Xu KM,
actual diagnosis. Slagboom PE, van Zwet EW, Lumey LH, Heijmans BT: DNA
methylation as a mediator of the association between prenatal
adversity and risk factors for metabolic disease in adulthood.
Sci Adv 2018, 4 eaao4364.
Conclusion
The recent interest in epigenetics, from a developmental 7. Feinberg AP: The key role of epigenetics in human disease
prevention and mitigation. N Engl J Med 2018, 378:1323-1334.
perspective, stems from the potential of DNA methyla-
8. Rakyan VK, Down TA, Balding DJ, Beck S: Epigenome-wide
tion to index both exposure to adversity and vulnerability  association studies for common human diseases. Nat Rev
for mental health problems [2]. To this end, there has Genet 2011, 12:529.
Overview of the strengths and, 2018 overview of the strengths and
been substantial activity in examining EWASes of adver- challenges of epigenetic research for complex diseases, with recom-
sity-related disorders, such as conduct disorder [37] and mendations for study design and other considerations.
psychosis [38]. Of interest, from these EWAS, DNAm in 9. Kitsios GD, Zintzaras E: Genome-wide association studies:
genes that underlie stress response, neurotransmitter hypothesis-“free” or “engaged”? Transl Res 2009, 154:161-164.
activity and immune regulation have been identified. 10. McCarthy MI, Abecasis GR, Cardon LR, Goldstein DB, Little J,
These preliminary findings may provide a useful frame- Ioannidis JP, Hirschhorn JN: Genome-wide association studies
for complex traits: consensus, uncertainty and challenges. Nat
work for more in-depth investigations — potentially as Rev Genet 2008, 9:356.
CNS surrogates or biomarkers — of the biological patho-
11. Gaunt TR, Shihab HA, Hemani G, Min JL, Woodward G,
genesis of a mental health problem. However, we argue  Lyttleton O, Zheng J, Duggirala A, McArdle WL, Ho K et al.:
that understanding hidden hypotheses within the EWAS Systematic identification of genetic influences on methylation
across the human life course. Genome Biol 2016, 17:1.
is an important first step in interpreting the results in Extremely well-done research studying genetic influences on DNA
relation to mental health phenotypes. methylation.
12. Pidsley R, Zotenko E, Peters TJ, Lawrence MG, Risbridger GP,
 Molloy P, Van Djik S, Muhlhausler B, Stirzaker C, Clark SJ: Critical
Conflict of interest statement evaluation of the Illumina MethylationEPIC BeadChip
Nothing declared. microarray for whole-genome DNA methylation profiling.
Genome Biol 2016, 17:208.
Rigorous comparison of the EPIC BeadChip to the HM450 BeadChip and
Whole Genome Sequencing, to evaluate agreement, coverage and
Acknowledgement content.
This manuscript was supported by a grant from the Economic and Social
Research Council (ES/R005516/2) to EDB. 13. Lin D, Chen J, Perrone-Bizzozero N, Bustillo JR, Du Y, Calhoun VD,
Liu J: Characterization of cross-tissue genetic-epigenetic
effects and their patterns in schizophrenia. Genome Med 2018,
10:13.
References and recommended reading
Papers of particular interest, published within the period of review, 14. Jones MJ, Moore SR, Kobor MS: Principles and challenges of
have been highlighted as: applying epigenetic epidemiology to psychology. Annu Rev
Psychol 2018, 69:459-485.
 of special interest
 of outstanding interest 15. Bibikova M, Barnes B, Tsan C, Ho V, Klotzle B, Le JM, Delano D,
Zhang L, Schroth GP, Gunderson KL et al.: High density DNA
methylation array with single CpG site resolution. Genomics
1. Shonkoff JP, Boyce WT, McEwen BS: Neuroscience, molecular 2011, 98:288-295.
biology, and the childhood roots of health disparities: building
a new framework for health promotion and disease 16. Bakulski KM, Halladay A, Hu VW, Mill J, Fallin MD: Epigenetic
prevention. J Am Med Assoc 2009, 301:2252-2259 http://dx.doi.  research in neuropsychiatric disorders: the “tissue issue”.
org/10.1001/jama.2009.754. Curr Behav Neurosci Rep 2016, 3:264-274.

Current Opinion in Psychology 2019, 27:13–17 www.sciencedirect.com

 Hidden hypotheses Barker, Roberts and Walton 17

Very good review on the challenges but also potential avenues of using 27. Felger JC, Li Z, Haroon E, Woolwine BJ, Jung MY, Hu X, Miller AH:
blood tissue in neuropsychiatric research. Inflammation is associated with decreased functional
connectivity within corticostriatal reward circuitry in
17. Walton E, Hass J, Liu J, Roffman JL, Bernardoni F, Roessner V, depression. Mol Psychiatry 2016, 21:1358.
Kirsch M, Schackert G, Calhoun V, Ehrlich S: Correspondence of
DNA methylation between blood and brain tissue and its 28. Barker ED, Cecil CAM, Walton E, Houtepen LC, O’Connor TG,
application to schizophrenia research. Schizophr Bull 2015.  Danese A, Jaffee SR, Jensen SKG, Pariante C, Roberts S,
sbv074. McArdle W, Gaunt TR, Relton CL: Inflammation-related
epigenetic risk and child and adolescent mental health: a
18. Ladd-Acosta C, Fallin MD: The role of epigenetics in genetic and prospective study from pregnancy to mid-adolescence. Dev
environmental epidemiology. Epigenomics 2015, 8:271-283 Psychopathol 2018. (in press).
http://dx.doi.org/10.2217/epi.15.102. First longitudinal study to employ inflammation-related poly epigenetic
genetic risk scores in relation to pre- and post-natal adversities, neuro-
19. O’Bryant SE, Mielke MM, Rissman RA, Lista S, Vanderstichele H, cognitive function and mental health problems.
Zetterberg H, Lewczuk P, Posner H, Hall J, Johnson L: Blood-
based biomarkers in Alzheimer disease: current state of the 29. Ladd-Acosta C: Epigenetic signatures as biomarkers of
science and a novel collaborative paradigm for advancing exposure. Curr Environ Health Rep 2015, 2:117-125.
from discovery to clinic. Alzheimer’s Dement 2017, 13:45-58.
30. Liu C, Marioni RE, Hedman AK, Pfeiffer L, Tsai PC, Reynolds LM,
20. Barker ED, Cecil CAM, Walton E, Meehan AJ: Genetic influences Just AC, Duan Q, Boer CG, Tanaka T et al.: A DNA methylation
on childhood disruptive behaviors. In The Wiley Handbook of biomarker of alcohol consumption. Mol Psychiatry 2016 http://
Disruptive and Impulse-Control Disorders. Edited by Lochman J, dx.doi.org/10.1038/mp.2016.192.
Matthys W. USA: Wiley-Blackwell; 2017.
31. Hattab M, Clark S, van den Oord E: Overestimation of the
21. Breton CV, Marsit CJ, Faustman E, Nadeau K, Goodrich JM, classification accuracy of a biomarker for assessing heavy
 Dolinoy DC, Herbstman J, Holland N, LaSalle JM, Schmidt R et al.: alcohol use. Mol Psychiatry 2017 http://dx.doi.org/10.1038/
Small-magnitude effect sizes in epigenetic end points are mp.2017.181.
important in children’s environmental health studies: the 32. Barker ED: Epigenetics, early adversity and child and
Children’s Environmental Health and Disease Prevention adolescent mental health. Psychopathology 2018, 51:71-75.
Research Center’s Epigenetics Working Group. Environ Health
Perspect 2017, 125:511. 33. Felix JF, Joubert BR, Baccarelli AA, Sharp GC, Almqvist C, Annesi-
Thorough review discussing methodologies in epigenetic research,  Maesano I, Arshad H, Baı̈z N, Bakermans-Kranenburg MJ,
robust findings (even of small effect sizes) in longitudinal work, and Bakulski KM et al.: Cohort profile: Pregnancy And Childhood
implications for the future. Epigenetics (PACE) Consortium. Int J Epidemiol 2017.
Important example of consortia type efforts.
22. Michels KB, Binder AM, Dedeurwaerder S, Epstein CB, Greally JM,
Gut I, Houseman EA, Izzi B, Kelsey KT, Meissner A, Milosavljevic A, 34. Hannon E, Dempster E, Viana J, Burrage J, Smith AR,
Siegmund KD, Bock C, Irizarry RA: Recommendations for the Macdonald R, St Clair D, Mustard C, Breen G, Therman S et al.: An
design and analysis of epigenome-wide association studies. integrated genetic-epigenetic analysis of schizophrenia:
Nat Methods 2013, 10:949-955 http://dx.doi.org/10.1038/ evidence for co-localization of genetic associations and
nmeth.2632. differential DNA methylation. Genome Biol 2016, 17:176.
23. Hannon E, Lunnon K, Schalkwyk L, Mill J: Interindividual 35. Joubert BR, Felix JF, Yousefi P, Bakulski KM, Just AC, Breton C,
 methylomic variation across blood, cortex, and cerebellum: Reese SE, Markunas CA, Richmond RC, Xu C-J: DNA
implications for epigenetic studies of neurological and methylation in newborns and maternal smoking in pregnancy:
neuropsychiatric phenotypes. Epigenetics 2015, 10:1024-1032. genome-wide consortium meta-analysis. Am J Hum Genet
This is an important paper, one of the first to show concordance in DNAm 2016, 98:680-696.
across tissue types, particularly blood and brain.
36. Insel T, Cuthbert B, Garvey M, Heinssen R, Pine DS, Quinn K,
24. Walton E, Cecil CAM, Suderman M, Liu J, Turner JA, Calhoun VD, Sanislow C, Wang P: Research domain criteria (RDoC): toward
Ehrlich S, Relton CL, Barker ED: Longitudinal epigenetic a new classification framework for research on mental
predictors of amygdala:hippocampus volume ratio. J Child disorders. Am Psychiatric Assoc 2010, 167:748-751.
Psychol Psychiatry 2017, 58:1341-1350.
37. Cecil CAM, Walton E, Jaffee SR, O’Connor T, Maughan B,
25. Hannon E, Weedon M, Bray N, O’Donovan M, Mill J: Pleiotropic Relton CL, Smith RG, McArdle W, Gaunt TR, Ouellet-Morin I,
effects of trait-associated genetic variation on DNA Barker ED: Neonatal DNA methylation and early-onset conduct
methylation: utility for refining GWAS loci. Am J Hum Genet problems: a genome-wide, prospective study. Dev
2017, 100:954-959. Psychopathol 2018, 30:383-397.
38. Fisher HL, Murphy TM, Arseneault L, Caspi A, Moffitt TE, Viana J,
26. Nusslock R, Miller GE: Early-life adversity and physical and
Hannon E, Pidsley R, Burrage J, Dempster EL et al.: Methylomic
emotional health across the lifespan: a neuroimmune network
analysis of monozygotic twins discordant for childhood
hypothesis. Biol Psychiatry 2016, 80:23-32.
psychotic symptoms. Epigenetics 2015, 10:1014-1023.

www.sciencedirect.com Current Opinion in Psychology 2019, 27:13–17