Brain age prediction using deep learning uncovers associated sequence variants

Machine learning algorithms can be trained to estimate age from brain structural MRI. The difference between an individual’s predicted and chronological age, predicted age difference (PAD), is a phenotype of relevance to aging and brain disease. Here, we present a new deep learning approach to predict brain age from a T1-weighted MRI. The method was trained on a dataset of healthy Icelanders and tested on two datasets, IXI and UK Biobank, utilizing transfer learning to improve accuracy on new sites. A genome-wide association study (GWAS) of PAD in the UK Biobank data (discovery set: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N=12378$$\end{document}N=12378, replication set: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N=4456$$\end{document}N=4456) yielded two sequence variants, rs1452628-T (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta =-0.08$$\end{document}β=−0.08, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P=1.15\times{10}^{-9}$$\end{document}P=1.15×10−9) and rs2435204-G (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta =0.102$$\end{document}β=0.102, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P=9.73\times 1{0}^{-12}$$\end{document}P=9.73×10−12). The former is near KCNK2 and correlates with reduced sulcal width, whereas the latter correlates with reduced white matter surface area and tags a well-known inversion at 17q21.31 (H2).

Machine learning algorithms can be trained to estimate age from brain structural MRI. Here, the authors introduce a new deep-learning-based age prediction approach, and then carry out a GWAS of the difference between predicted and chronological age, revealing two associated variants.

Rare and Common Variants Conferring Risk of Tooth Agenesis

We present association results from a large genome-wide association study of tooth agenesis (TA) as well as selective TA, including 1,944 subjects with congenitally missing teeth, excluding third molars, and 338,554 controls, all of European ancestry. We also tested the association of previously identified risk variants, for timing of tooth eruption and orofacial clefts, with TA. We report associations between TA and 9 novel risk variants. Five of these variants associate with selective TA, including a variant conferring risk of orofacial clefts. These results contribute to a deeper understanding of the genetic architecture of tooth development and disease. The few variants previously associated with TA were uncovered through candidate gene studies guided by mouse knockouts. Knowing the etiology and clinical features of TA is important for planning oral rehabilitation that often involves an interdisciplinary approach.

Genome-wide common and rare variant analysis provides novel insights into clozapine-associated neutropenia

This corrects the article DOI: 10.1038/mp.2016.97.

15q11.2 CNV affects cognitive, structural and functional correlates of dyslexia and dyscalculia

Several copy number variants have been associated with neuropsychiatric disorders and these variants have been shown to also influence cognitive abilities in carriers unaffected by psychiatric disorders. Previously, we associated the 15q11.2(BP1-BP2) deletion with specific learning disabilities and a larger corpus callosum. Here we investigate, in a much larger sample, the effect of the 15q11.2(BP1-BP2) deletion on cognitive, structural and functional correlates of dyslexia and dyscalculia. We report that the deletion confers greatest risk of the combined phenotype of dyslexia and dyscalculia. We also show that the deletion associates with a smaller left fusiform gyrus. Moreover, tailored functional magnetic resonance imaging experiments using phonological lexical decision and multiplication verification tasks demonstrate altered activation in the left fusiform and the left angular gyri in carriers. Thus, by using convergent evidence from neuropsychological testing, and structural and functional neuroimaging, we show that the 15q11.2(BP1-BP2) deletion affects cognitive, structural and functional correlates of both dyslexia and dyscalculia.

Genome-wide common and rare variant analysis provides novel insights into clozapine-associated neutropenia

This corrects the article DOI: 10.1038/mp.2016.97.

De novo CNV analysis implicates specific abnormalities of postsynaptic signalling complexes in the pathogenesis of schizophrenia

A small number of rare, recurrent genomic copy number variants (CNVs) are known to substantially increase susceptibility to schizophrenia. As a consequence of the low fecundity in people with schizophrenia and other neurodevelopmental phenotypes to which these CNVs contribute, CNVs with large effects on risk are likely to be rapidly removed from the population by natural selection. Accordingly, such CNVs must frequently occur as recurrent de novo mutations. In a sample of 662 schizophrenia proband-parent trios, we found that rare de novo CNV mutations were significantly more frequent in cases (5.1% all cases, 5.5% family history negative) compared with 2.2% among 2623 controls, confirming the involvement of de novo CNVs in the pathogenesis of schizophrenia. Eight de novo CNVs occurred at four known schizophrenia loci (3q29, 15q11.2, 15q13.3 and 16p11.2). De novo CNVs of known pathogenic significance in other genomic disorders were also observed, including deletion at the TAR (thrombocytopenia absent radius) region on 1q21.1 and duplication at the WBS (Williams-Beuren syndrome) region at 7q11.23. Multiple de novos spanned genes encoding members of the DLG (discs large) family of membrane-associated guanylate kinases (MAGUKs) that are components of the postsynaptic density (PSD). Two de novos also affected EHMT1, a histone methyl transferase known to directly regulate DLG family members. Using a systems biology approach and merging novel CNV and proteomics data sets, systematic analysis of synaptic protein complexes showed that, compared with control CNVs, case de novos were significantly enriched for the PSD proteome (P=1.72 × 10⁻⁶. This was largely explained by enrichment for members of the N-methyl-D-aspartate receptor (NMDAR) (P=4.24 × 10⁻⁶) and neuronal activity-regulated cytoskeleton-associated protein (ARC) (P=3.78 × 10⁻⁸) postsynaptic signalling complexes. In an analysis of 18 492 subjects (7907 cases and 10 585 controls), case CNVs were enriched for members of the NMDAR complex (P=0.0015) but not ARC (P=0.14). Our data indicate that defects in NMDAR postsynaptic signalling and, possibly, ARC complexes, which are known to be important in synaptic plasticity and cognition, play a significant role in the pathogenesis of schizophrenia.

Copy number variations of chromosome 16p13.1 region associated with schizophrenia

Deletions and reciprocal duplications of the chromosome 16p13.1 region have recently been reported in several cases of autism and mental retardation (MR). As genomic copy number variants found in these two disorders may also associate with schizophrenia, we examined 4345 schizophrenia patients and 35,079 controls from 8 European populations for duplications and deletions at the 16p13.1 locus, using microarray data. We found a threefold excess of duplications and deletions in schizophrenia cases compared with controls, with duplications present in 0.30% of cases versus 0.09% of controls (P=0.007) and deletions in 0.12 % of cases and 0.04% of controls (P>0.05). The region can be divided into three intervals defined by flanking low copy repeats. Duplications spanning intervals I and II showed the most significant (P = 0.00010) association with schizophrenia. The age of onset in duplication and deletion carriers among cases ranged from 12 to 35 years, and the majority were males with a family history of psychiatric disorders. In a single Icelandic family, a duplication spanning intervals I and II was present in two cases of schizophrenia, and individual cases of alcoholism, attention deficit hyperactivity disorder and dyslexia. Candidate genes in the region include NTAN1 and NDE1. We conclude that duplications and perhaps also deletions of chromosome 16p13.1, previously reported to be associated with autism and MR, also confer risk of schizophrenia.

Screen for CACNA1A and ATP1A2 mutations in sporadic hemiplegic migraine patients

The aim of this study was to investigate the involvement of the CACNA1A and ATP1A2 gene in a population-based sample of sporadic hemiplegic migraine (SHM). Patients with SHM (n = 105) were identified in a nationwide search in the Danish population. We sequenced all exons and promoter regions of the CACNA1A and ATP1A2 genes in 100 patients with SHM to search for possible SHM mutations. Novel DNA variants were discovered in eight SHM patients, four in exons of the CACNA1A gene and four in exons of the ATP1A2 gene. Six of the variants were considered non-pathogenic. The causal role of the two remaining DNA variants is unknown until functional studies have been made or independent genetic evidence is discovered. Only very few DNA variants were identified in 100 SHM patients, and regardless of whether the identified variants are causal the CACNA1A and ATP1A2 genes are not major genes in SHM.

The genetic spectrum of a population-based sample of familial hemiplegic migraine

Familial hemiplegic migraine (FHM) is a rare subtype of migraine with aura and transient hemiplegia. FHM mutations are known in three genes, the CACNA1A (FHM1) gene, the ATP1A2 (FHM2) and the SCN1A (FHM3) gene and seem to have an autosomal-dominant mode of inheritance. The aim of this study was to search for FHM mutations in FHM families identified through a screen of the Danish population of 5.2 million people. FHM patients were diagnosed according to the International Classification of Headache Disorders and all FHM patients had a physical and neurological examination by a physician. A total of 147 FHM patients from 44 different families were identified; 43 FHM families participated in this study. Linkage analysis of these families shows clear linkage to the FHM locus (FHM1) on chromosome 19, supportive linkage to the FHM2 locus whereas no linkage was found to the FHM3 locus. Furthermore, we sequenced all exons and promoter regions of the CACNA1A and ATP1A2 genes and screened for the Q1489K mutation in the SCN1A gene. CACNA1A gene mutations were identified in three of the FHM families, two known FHM mutations, R583Q and T666M and one novel C1369Y mutation. Three FHM families were identified with novel mutations in the ATP1A2 gene; a family with a V138A mutation, a family with a R202Q mutation and a family with a R763C mutation. None of the Danish FHM families have the Q1489K mutation in the SCN1A gene. Our study shows that only 14% (6/42) of FHM families in the general Danish population have exonic FHM mutations in the CACNA1A or ATP1A2 gene. The families we identified with FHM mutations in the CACNA1A and ATP1A2 genes were extended, multiple affected families whereas the remaining FHM families were smaller. The existence of many small families in the Danish FHM cohort may reflect less bias in FHM family ascertainment and/or more locus heterogeneity than described previously.

Identification of a novel neuregulin 1 at-risk haplotype in Han schizophrenia Chinese patients, but no association with the Icelandic/Scottish risk haplotype

To determine if neuregulin 1 (NRG1) is associated with schizophrenia in Asian populations, we investigated a Han Chinese population using both a family trio design and a case-control design. A total of 25 microsatellite markers and single nucleotide polymorphisms (SNPs) were genotyped spanning the 1.1 Mb NRG1 gene including markers of a seven-marker haplotype at the 5' end of the gene found to be in excess in Icelandic and Scottish schizophrenia patients. The alleles of the individual markers forming the seven marker at-risk haplotype are not likely to be causative as they are not in excess in patients in the Chinese population studied here. However using unrelated patients, we find a novel haplotype (HAP(China 1)), immediately upstream of the Icelandic haplotype, in excess in patients (11.9% in patients vs 4.2% in controls; P=0.0000065, risk ratio (rr) 3.1), which was not significant when parental controls were used. Another haplotype (HAP(China 2)) overlapping the Icelandic risk haplotype was found in excess in the Chinese (8.5% of patients vs 4.0% of unrelated controls; P=0.003, rr 2.2) and was also significant using parental controls only (P=0.0047, rr 2.1). A four-marker haplotype at the 3' end of the NRG1 gene, HAP(China 3), was found at a frequency of 23.8% in patients and 13.7% in nontransmitted parental haplotypes (P=0.000042, rr=2.0) but was not significant in the case-control comparison. We conclude that different haplotypes within the boundaries of the NRG1 gene may be associated with schizophrenia in the Han Chinese.

Genetic factors contribute to the risk of developing endometriosis

BACKGROUND

Endometriosis is known to cluster within nuclear families. The extent of familial clustering can be evaluated in Iceland with its large population-based genealogical database.

METHODS AND RESULTS

Applying several measures of familiality we demonstrated that 750 women with endometriosis were significantly more interrelated than matched control groups. The risk ratio for sisters was 5.20 (P < 0.001) and for cousins 1.56 (P = 0.003). The average kinship coefficient for the patients was significantly higher than that calculated for 1000 sets of 750 matched controls (P < 0.001) and this remained significant when contribution from first-degree relatives was excluded (P < 0.05). The minimum number of ancestors required to account for the group of patients was compared with the minimum number of ancestors required to account for the control groups at different time points in the past. The minimum number of founders for the group of patients was significantly smaller than for the control groups. Affected cousin pairs were as likely to be paternally connected as maternally connected.

CONCLUSIONS

This is the first population-based study using an extensive genealogy database to examine the genetic contribution to endometriosis. A genetic factor is present, with a raised risk in close and more distant relatives, and a definite kinship factor with maternal and paternal inheritance contributing.

Endometriosis is not associated with or linked to the GALT gene

OBJECTIVE

To investigate a possible association between the carrier frequency of the N314D mutation in the galactose-1-phosphate uridyl transferase (GALT) gene and endometriosis and linkage to the short arm of chromosome 9, where the GALT gene resides.

DESIGN

Association and linkage study.

SETTING

Population material collected for case and family studies in endometriosis.

PATIENT(S)

Women diagnosed with endometriosis by laparotomy or laparoscopy.

INTERVENTION(S)

Association with the GALT gene investigated by genotyping 85 affected women and 213 unrelated control women and a scan for linkage to chromosome 9 in 205 women from 64 families with endometriosis.

MAIN OUTCOME MEASURE(S)

Multipoint parametric lod scores and frequency of alleles.

RESULT(S)

There was no significant difference in allele frequency for the N314D polymorphism in patients compared with control subjects. No evidence for linkage was found to chromosome 9p, where the GALT gene resides.

CONCLUSION(S)

The experiments reported herein provide no evidence supporting involvement of the GALT locus in the development of endometriosis.

Photosystem II in different parts of the thylakoid membrane: a functional comparison between different domains

The electron transport properties of photosystem II (PSII) from five different domains of the thylakoid membrane were analyzed by flash-induced fluorescence kinetics. These domains are the entire grana, the grana core, the margins from the grana, the stroma lamellae, and the Y100 fraction (which represent more purified stroma lamellae). The two first fractions originate from appressed grana membranes and have PSII with a high proportion of O(2)-evolving centers (80-90%) and efficient electron transport on the acceptor side. About 30% of the granal PSII centers were found in the margin fraction. Two-thirds of those PSII centers evolve O(2), but the electron transfer on the acceptor side is slowed. PSII from the stroma lamellae was less active. The fraction containing the entire stroma has only 43% O(2)-evolving PSII centers and slow electron transfer on the acceptor side. In contrast, PSII centers of the Y100 fraction show no O(2) evolution and were unable to reduce Q(B). Flash-induced fluorescence decay measurements in the presence of DCMU give information about the integrity of the donor side of PSII. We were able to distinguish between PSII centers with a functional Mn cluster and without any Mn cluster, and PSII centers which undergo photoactivation and have a partially assembled Mn cluster. From this analysis, we propose the existence of a PSII activity gradient in the thylakoid membrane. The gradient is directed from the stroma lamellae, where the Mn cluster is absent or inactive, via the margins where photoactivation accelerates, to the grana core domain where PSII is fully photoactivated. The photoactivation process correlates to the PSII diffusion along the membrane and is initiated in the stroma lamellae while the final steps take place in the appressed regions of the grana core. The margin domain is seemingly very important in this process.

Two dimensional electrophoresis of thylakoid membrane proteins and its application to microsequencing

A procedure of two-dimensional gel electrophoresis adapted for application on membrane proteins from the thylakoids is described. It involves isoelectric focusing in the first dimension and size dependent electrophoresis in the second dimension. About 100 polypeptides are clearly separated with relatively little streaking. About 20 polypeptides are identified by immunoblotting or location in the gel. They are the polypeptides of the PS I core, the 64 kDa protein, the α and β subunits of CF1 ATPase, cytochrome f, Rieske iron-sulfur protein, the 23 kDa and 33 kDa polypeptides of the oxygen evolving complexes, CP29, CP24, CP27 and CP25 (last two proteins belong to LHCII). Some proteins give rise to two or more separate spots indicating a separation of different isoforms of these proteins. Among them, the LHCII polypeptides (27 kDa and 25 kDa) were each resolved into at least three spots in the pH range 4.75-5.90; the Rieske FeS protein, as published elsewhere (Yu et al. 1994), was separated into two forms having different isoelectric points (pI 5.1 and 5.4), each of them was also microsequenced; the 64 kDa protein claimed to be a LHCII-kinase was found to be multiple forms appearing in at least two isoforms with pI 6.2 (K1) and 6.0 (K2) respectively, furthermore, K1 can be resolved into two subpopulations.The lateral distribution of these proteins in the thylakoid membrane was determined by analysing the vesicles originating from different parts of the thylakoids. The data obtained from this analysis can be partially used as markers for different thylakoid domains.This procedure for sample solubilization and 2-D electrophoresis is useful for the analysis of the polypeptide composition of vesicles originating from the thylakoid membrane and for microsequences of individual polypeptides isolated from the 2-D gel.