Search

Detecting epistatic drivers of human phenotypes remains a challenge. Traditional approaches use regression to sequentially test multiplicative interaction terms involving single pairs of genetic variants. For higher-order interactions and genome-wide large-scale data, this strategy is computationally intractable. Moreover, multiplicative terms used in regression modeling may not capture the form of biological interactions. Building on the Predictability, Computability, Stability (PCS) framework, we introduce the epiTree pipeline to extract higher-order interactions from genomic data using tree-based models. The epiTree pipeline first selects a set of variants derived from tissue-specific estimates of gene expression. Next, it uses iterative random forests (iRF) to search training data for candidate Boolean interactions (pairwise and higher-order). We derive significance tests from interactions by simulating Boolean tree-structured null (no epistasis) and alternative (epistasis) distributions on hold-out test data. Finally, our pipeline computes PCS epistasis p-values that evaluate the stability of improvement in prediction accuracy via bootstrap sampling on the test set. We validate the epiTree pipeline using the phenotype of red-hair from the UK Biobank, where several genes are known to demonstrate epistatic interactions. epiTree recovers both previously reported and novel interactions, which represent forms of non-linearities not captured by logistic regression models. Additionally, epiTree suggests interactions between genes such as PKHD1 and XPOTP1, which are unlinked to MC1R, as novel candidate interactions associated with the red hair phenotype. Last but not least, we find that individual Boolean or tree-based epistasis models generally provide higher prediction accuracy than classical logistic regression.

The aortic valve is an important determinant of cardiovascular physiology and anatomic location of common human diseases. From a sample of 26,142 European-ancestry participants, we estimated functional aortic valve area by planimetry from prospectively obtained cardiac MRI sequences of the aortic valve. A genome-wide association study of aortic valve area in these UK Biobank participants showed two significant associations indexed by rs71190365 (chr13:50764607,DLEU1, p=1.8×10−9) and rs35991305 (chr12:94191968,CRADD, p=3.4×10−8). From the GWAS findings we constructed a polygenic risk score for aortic valve area, which in a separate cohort of 311,728 individuals without imaging demonstrated that smaller aortic valve area is predictive of increased risk for aortic valve disease (Odds Ratio 0.88,p=2.3×10−6). After excluding subjects with a medical diagnosis of aortic valve stenosis (remaining n=310,546 individuals), phenome-wide association of >10,000 traits showed multiple links between the polygenic score for aortic valve disease and key health-related comorbidities involving the cardiovascular system and autoimmune disease. Genetic correlation analysis supports a shared genetic etiology with between aortic valve size and birthweight along with other cardiovascular conditions. These results illustrate the use of automated phenotyping of cardiac imaging data from the general population to investigate the genetic etiology of aortic valve disease, perform clinical prediction, and uncover new clinical and genetic correlates of cardiac anatomy.

SUMMARYMendelian Randomization (MR) is an important causal inference method primarily used in biomedical research. This work applies contemporary techniques in machine learning to improve the robustness and power of traditional MR tools. By denoising and combining candidate genetic variants through techniques from unsupervised probabilistic graphical models, an influential latent instrumental variable is constructed for causal effect estimation. We present results on identifying relationships between biomarkers and the occurrence of coronary artery disease using individual-level real-world data from UK-BioBank via the proposed method. The approach, termed Instrumental Variable sYnthesis (IVY) is proposed as a complement to current methods, and is able to improve results based on allele scoring, particularly at moderate sample sizes.

Sensitivity to external demands is essential for adaptation to dynamic environments, but comes at the cost of increased risk of adverse outcomes when facing poor environmental conditions. Here, we apply a novel methodology to perform genome-wide association analysis of mean and variance in ten key brain features (accumbens, amygdala, caudate, hippocampus, pallidum, putamen, thalamus, intracranial volume, cortical surface area, and cortical thickness), integrating genetic and neuroanatomical data from a large lifespan sample (n = 25,575 individuals; 8–89 years, mean age 51.9 years). We identify genetic loci associated with phenotypic variability in thalamus volume and cortical thickness. The variance-controlling loci involved genes with a documented role in brain and mental health and were not associated with the mean anatomical volumes. This proof-of-principle of the hypothesis of a genetic regulation of brain volume variability contributes to establishing the genetic basis of phenotypic variance (i.e., heritability), allows identifying different degrees of brain robustness across individuals, and opens new research avenues in the search for mechanisms controlling brain and mental health. This is an Accepted Manuscript of an article published by Nature Publishing Group, available online: http://dx.doi.org/10.1038/s41380-020-0664-1

ImportanceBetween-subject variability in brain structure is determined by gene-environment interactions, possibly reflecting differential sensitivity to environmental and genetic perturbations. Magnetic resonance imaging (MRI) studies have revealed thinner cortices and smaller subcortical volumes in patients. However, such group-level comparisons may mask considerable within-group heterogeneity, which has largely remained unnoticed in the literatureObjectiveTo compare brain structural variability between individuals with SZ and healthy controls (HC) and to test if respective variability reflects the polygenic risk for SZ (PRS) in HC.Design, Setting, and ParticipantsWe compared MRI derived cortical thickness and subcortical volumes between 2,010 healthy controls and 1,151 patients with SZ across 16 cohorts. Secondly, we tested for associations between PRS and MRI features in 12,490 participants from UK Biobank.Main Outcomes and MeasuresWe modeled mean and dispersion effects of SZ and PRS using double generalized linear models. We performed vertex-wise analyses for thickness, and region-of-interest analysis for cortical, subcortical and hippocampal subfield volumes. Follow-up analyses included within-sample analysis, controlling for intracranial volume and population covariates, test of robustness of PRS threshold, and outlier removal.ResultsCompared to controls, patients with SZ showed higher heterogeneity in cortical thickness, cortical and ventricle volumes, and hippocampal subfields. Higher PRS was associated with thinner frontal and temporal cortices, as well as smaller left CA2/3, but was not significantly associated with dispersion.Conclusion and relevanceSZ is associated with substantial brain structural heterogeneity beyond the mean differences. These findings possibly reflect higher differential sensitivity to environmental and genetic perturbations in patients, supporting the heterogeneous nature of SZ. Higher PRS for SZ was associated with thinner fronto-temporal cortices and smaller subcortical volumes, but there were no significant associations with the heterogeneity in these measures, i.e. the variability among individuals with high PRS were comparable to the variability among individuals with low PRS. This suggests that brain variability in SZ results from interactions between environmental and genetic factors that are not captured by the PGR. Factors contributing to heterogeneity in fronto-temporal cortices and hippocampus are thus key to further our understanding of how genetic and environmental factors shape brain biology in SZ.Key PointsQuestion:Is schizophrenia and its polygenic risk associated with brain structural heterogeneity in addition to mean changes?Findings:In a sample of 1151 patients and 2010 controls, schizophrenia was associated with increased heterogeneity in fronto-temporal thickness, cortical, ventricle, and hippocampal volumes, besides robust reductions in mean estimates. In an independent sample of 12,490 controls, polygenic risk for schizophrenia was associated with thinner fronto-temporal cortices and smaller CA2/3 of the left hippocampus, but not with heterogeneity.Meaning:Schizophrenia is associated with increased inter-individual differences in brainstructure, possibly reflecting clinical heterogeneity, gene-environment interactions, or secondary disease factors.

Supported by histological and genetic evidence implicating myelin, neuroinflammation and oligodendrocyte dysfunction in schizophrenia spectrum disorders (SZ), diffusion tensor imaging (DTI) studies have consistently shown white matter (WM) abnormalities when compared to healthy controls (HC). The diagnostic specificity remains unclear, with bipolar disorders (BD) frequently conceptualized as a less severe clinical manifestation along a psychotic spectrum. Further, the age-related dynamics and possible sex differences of WM abnormalities in SZ and BD are currently understudied.Using tract-based spatial statistics (TBSS) we compared DTI-based microstructural indices between SZ (n=128), BD (n=61), and HC (n=293). We tested for age-by-group and sex-by-group interactions, computed effect sizes within different age-bins and within genders.TBSS revealed global reductions in fractional anisotropy (FA) and increases in radial (RD) diffusivity in SZ compared to HC, with strongest effects in the body and splenium of the corpus callosum, and lower FA in SZ compared to BD in right inferior longitudinal fasciculus and right inferior fronto-occipital fasciculus, and no significant differences between BD and HC. The results were not strongly dependent on age or sex. Despite lack of significant group-by-age interactions, a sliding-window approach supported widespread WM involvement in SZ with most profound differences in FA from the late 20s.

Numerous genetic and environmental factors contribute to psychiatric disorders and other brain disorders. Common risk factors likely converge on biological pathways regulating the optimization of brain structure and function across the lifespan. Here, using structural magnetic resonance imaging and machine learning, we estimated the gap between brain age and chronological age in 36,891 individuals aged 3 to 96 years, including individuals with different brain disorders. We show that several disorders are associated with accentuated brain aging, with strongest effects in schizophrenia, multiple sclerosis and dementia, and document differential regional patterns of brain age gaps between disorders. In 16,269 healthy adult individuals, we show that brain age gap is heritable with a polygenic architecture overlapping those observed in common brain disorders. Our results identify brain age gap as a genetically modulated trait that offers a window into shared and distinct mechanisms in different brain disorders.

The hippocampus is a heterogeneous structure, comprising histologically distinguishable subfields. These subfields are differentially involved in memory consolidation, spatial navigation and pattern separation, complex functions often impaired in individuals with brain disorders characterized by reduced hippocampal volume, including Alzheimer’s disease (AD) and schizophrenia. Given the structural and functional heterogeneity of the hippocampal formation, we sought to characterize the subfields’ genetic architecture. T1-weighted brain scans (n=21297, 16 cohorts) were processed with the hippocampal subfields algorithm in FreeSurfer v6.0. We ran a genome-wide association analysis on each subfield, covarying for total hippocampal volume. We further calculated the single nucleotide polymorphism (SNP)-based heritability of twelve subfields, as well as their genetic correlation with each other, with other structural brain features, and with AD and schizophrenia. All outcome measures were corrected for age, sex, and intracranial volume. We found 15 unique genome-wide significant loci across six subfields, of which eight had not been previously linked to the hippocampus. Top SNPs were mapped to genes associated with neuronal differentiation, locomotor behaviour, schizophrenia and AD. The volumes of all the subfields were estimated to be heritable (h2 from .14 to .27, all p< 1×10-16) and clustered together based on their genetic correlations compared to other structural brain features. There was also evidence of genetic overlap of subicular subfield volumes with schizophrenia. We conclude that hippocampal subfields have partly distinct genetic determinants associated with specific biological processes and traits. Taking into account this specificity may increase our understanding of hippocampal neurobiology and associated pathologies.

ImportanceAccumulating evidence supports cerebellar involvement in mental disorders such as schizophrenia, bipolar disorder, depression, anxiety disorders and attention-deficit hyperactivity disorder. However, little is known about cerebellar involvement in the developmental stages of these disorders. In particular, whether cerebellar morphology is associated with early expression of specific symptom domains remains unclear.ObjectiveTo determine the robustness and specificity of associations between cerebellar morphology, general cognitive function, general psychopathology and sub-clinical psychiatric symptom domains in adolescence.Design, setting and participantsAssessment of parametric structure-function associations between MR-based brain morphometric features and data-driven cognitive and clinical phenotypes in the Philadelphia Neurodevelopmental Cohort (N=1401, age-range: 8 - 23).Main outcomes and measuresRobust prediction of cognitive and clinical symptom domain scores from cerebellar, subcortical and cerebro-cortical brain features using machine learning with 10-fold internal cross-validation and permutation-based statistical inference.ResultsCerebellar morphology predicted both general cognitive function and general psychopathology (mean Pearson correlation coefficients between predicted and observed values: r = .20 and r = .13, respectively; corrected p-values < .0009). Analyses of specific sub-clinical symptom domains revealed significant associations with rates of norm-violating behavior (r = .17; p < .0009), prodromal psychotic symptoms (r = .12; p < .0009) and anxiety symptoms (r = .09; p =.0117). In contrast, we observed no significant associations between cerebellar features and the severity of attention deficits, depressive, manic or obsessive-compulsive symptoms (all rs =< .03, all ps => .1). Associations with norm-violating behavior and prodromal psychotic symptoms were stronger for the cerebellum than for subcortical and cerebro-cortical regions, while anxiety and general cognitive function were related to more global brain morphology patterns. The association between cerebellar volume and prodromal psychotic symptoms, and to a lesser extent norm violating behavior, remained significant when adjusting for potentially confounding factors such as general cognitive function, general psychopathology, parental education level and use of psychoactive substances.Conclusions and relevanceThe robust associations with sub-clinical psychiatric symptoms in the age range when these typically emerge highlight the cerebellum as a key brain structure in the development of severe mental disorders.Key pointsQuestionsIs cerebellar morphology associated with sub-clinical psychiatric symptoms in adolescence? Do such associations show symptom domain specificity or do they rather constitute a marker of general psychopathology?FindingsMachine learning utilizing cerebellar morphology features significantly predicted the severity of prodromal psychotic symptoms, norm-violating behavior and anxiety, but not attention deficits, depressive, manic or obsessive-compulsive sub-clinical symptoms. Associations with prodromal psychotic symptoms were stronger for the cerebellum than for cerebral subcortical and cerebro-cortical regions, and remained significant when adjusting for several potentially confounding factors.MeaningThe cerebellum appears to play a key role in the development of severe mental illness.

Loss of function variants in NOTCH1 cause left ventricular outflow tract obstructive defects (LVOTO) in a small percentage of families. Clinical surveys report an increased prevalence of missense variants in NOTCH1 in family members of individuals with LVOTO and other types of congenital heart disease (CHD). However, the risk conferred by rare variants in NOTCH1 for LVOTO remains largely uncharacterized. In a cohort of 49 families affected by hypoplastic left heart syndrome, a severe form of LVOTO, we discovered predicted loss of function NOTCH1 variants in 6% of individuals. Rare missense variants were found in an additional 16% of families. To make a quantitative estimate of the genetic risk posed by variants in NOTCH1 for LVOTO, we studied associations of 400 coding and non-coding variants in NOTCH1 in 271 adult cases and 333,571 controls from the UK Biobank. Two rare intronic variants in strong linkage disequilibrium displayed significant association with risk for LVOTO (g.chr9:139427582C>T, Odds Ratio 16.9, p=3.12e-6; g.chr9:139435649C>T, Odds Ratio 19.6, p = 2.44e-6) amongst European-ancestry British individuals. This result was replicated in an independent analysis of 51 cases and 68,901 controls of non-European and mixed ancestry. We conclude that carrying rare predicted loss of function variants or either of two intronic variants in NOTCH1 confer significant risk for LVOTO. Our approach demonstrates the utility of population-based datasets in quantifying the specific risk of individual variants for disease related phenotypes.Author summaryCongenital heart defects are the most common class of birth defect and are present in 1% of live births. Although CHD cases are often clustered in families, and thus the causal variant(s) are seemingly inherited, finding genetic variants causing these defects has been challenging. With the knowledge that variation in the NOTCH1 gene previously has been associated with CHDs affecting the left side of the heart, our aim was to further investigate the role of different types of NOTCH1 variants in left sided CHDs in two cohorts – a cohort of Finnish families with severe lesions affecting the left side of the heart, and the UK Biobank population including individuals with less severe left-sided lesions such as bicuspid aortic valve, congenital aortic stenosis, and coarctation of the aorta. We found a causal loss-of-function NOTCH1 variant in 6% of the families in the Finnish cohort and in the UK Biobank cohort, we identified two rare variants in the non-coding region of NOTCH1, associated with severe left-sided CHDs. These findings support screening of NOTCH1 loss-of-function variants in patients with severe left sided congenital heart defects and suggests that non-coding region variants in NOTCH1 play a role in CHDs.

A healthy transition from adolescence to adulthood relies on a continuous individual adaptation to a dynamic environment. Here, we employed data driven multivariate approaches to derive both general cognitive and psychopathology factors as well as brain phenotypes in children and adolescents in the publicly available PNC sample. We identified a distinct brain white matter pattern which proved central for prediction of heritable cognition and psychopathology scores, highlighting the importance of fronto-temporal connections for intellectual and mental development.

Cognitive and brain development are determined by dynamic interactions between genes and environment across the lifespan. Aside from marker-by-marker analyses of polymorphisms, biologically meaningful features of the whole-genome (derived from the combined effect of individual markers) have been postulated to inform on human phenotypes including cognitive traits and their underlying biological substrate.Here, estimates of inbreeding and genetic susceptibility for schizophrenia calculated from genome-wide data –runs of homozygosity (ROH) and schizophrenia polygenic risk score (PGRS)– are analyzed in relation to cognitive abilities (n=4183) and brain structure (n=516) in a general-population sample of European-ancestry participants aged 8-22, from the Philadelphia Neurodevelopmental Cohort.The findings suggest that a higher ROH burden and higher schizophrenia PGRS are associated with higher intelligence. Cognition~ROH and cognition~PGRS associations obtained in this cohort may respectively evidence that assortative mating influences intelligence, and that individuals with high schizophrenia genetic risk who do not transition to disease status are cognitively resilient.Neuroanatomical data showed that the effects of schizophrenia PGRS on cognition could be modulated by brain structure, although larger imaging datasets are needed to accurately disentangle the underlying neural mechanisms linking IQ with both inbreeding and the genetic burden for schizophrenia.