Polygenic Risk Scores (PRS) for AD offer unique possibilities for reliable identification of individuals at high and low risk of AD. However, there is little agreement in the field as to what approach should be used for genetic risk score calculations, how to model the effect of APOE, what the optimal p-value threshold (pT) for SNP selection is and how to compare scores between studies and methods. We show that the best prediction accuracy is achieved with a model with two predictors (APOE and PRS excluding APOE region) with pT<0.1 for SNP selection. Prediction accuracy in a sample across different PRS approaches is similar, but individuals’ scores and their associated ranking differ. We show that standardising PRS against the population mean, as opposed to the sample mean, makes the individuals’ scores comparable between studies. Our work highlights the best strategies for polygenic profiling when assessing individuals for AD risk.

Our study suggests that the clinical spectrum of early AD pathology is explained by different biological pathways, in particular, the endocytosis, clathrin/AP2 adaptor complex, and immune response pathways, that are independent of apolipoprotein E (APOE).

Genome-wide association studies have identified nearly 40 genome-wide significant single nucleotide polymorphisms (SNPs) which are associated with Alzheimer's Disease (AD). Due to the polygenicity of AD, polygenic risk scores (PRS) have shown high potential for AD risk prediction. PRSs have been shown to successfully discriminate between AD cases and controls achieving a prediction accuracy of up to 84% based on area under the receiver operating curve. The prediction accuracy in AD is higher compared with other complex genetic disorders. PRS can be restricted to SNPs which reside in biologically relevant gene-sets; the predictive value of these gene-sets in the general population is not as high as genome-wide PRS, but they may play an important role to identify mechanisms of disease development and inform biological experiments. Multiple methods are available to derive PRSs, such as selecting SNPs based on statistical evidence of association with the disease or using prior evidence for SNP selection. All methods have advantages, but PRS produced using different methodologies are often not comparable, and results should be interpreted with care. Similarly, this is true when PRS is based on different background populations. With the exponential growth in development of digital electronic devices it is easy to calculate an individual's disease risk using public databases. A major limitation for the utility of PRSs is that the risk score is sample and method dependent. Therefore, replicability and interpretability of PRS is an important issue. PRS can be used to determine the probability of developing disease which incorporates information about disease risk in the general population or in a specific AD risk group. It is essential to consult with genetic counselors to ensure genetic risk is communicated appropriately.

Alzheimer's disease (AD) is a devastating neurodegenerative condition with significant genetic heritability. Several genes have been implicated in the onset of AD with the apolipoprotein E ( APOE ) gene being the strongest single genetic risk loci. Evidence suggests that the effect of APOE alters with age during disease progression. Here, we aim to investigate the impact of APOE and other variants outside the APOE region on AD risk in younger and older participants. Using data from both the Alzheimer's Disease Neuroimaging Initiative and the UK Biobank, we computed the polygenic risk score of each individual informed by the latest genetic study from the International Genomics of Alzheimer's Project. Our analysis showed that the effect of APOE on the disease risk is greater in younger participants and reduces as participants' age increases. Our findings indicate the increased impact of polygenic risk score as participants' age increases. Therefore, AD in older individuals can potentially be triggered by the cumulative effect of genes which are outside the APOE region.

Polygenic risk scores (PRSs) are a method to summarize the additive trait variance captured by a set of SNPs, and can increase the power of set‐based analyses by leveraging public genome‐wide association study (GWAS) datasets. PRS aims to assess the genetic liability to some phenotype on the basis of polygenic risk for the same or different phenotype estimated from independent data. We propose the application of PRSs as a set‐based method with an additional component of adjustment for linkage disequilibrium (LD), with potential extension of the PRS approach to analyze biologically meaningful SNP sets. We call this method POLARIS: POlygenic Ld‐Adjusted RIsk Score. POLARIS identifies the LD structure of SNPs using spectral decomposition of the SNP correlation matrix and replaces the individuals' SNP allele counts with LD‐adjusted dosages. Using a raw genotype dataset together with SNP effect sizes from a second independent dataset, POLARIS can be used for set‐based analysis. MAGMA is an alternative set‐based approach employing principal component analysis to account for LD between markers in a raw genotype dataset. We used simulations, both with simple constructed and real LD‐structure, to compare the power of these methods. POLARIS shows more power than MAGMA applied to the raw genotype dataset only, but less or comparable power to combined analysis of both datasets. POLARIS has the advantages that it produces a risk score per person per set using all available SNPs, and aims to increase power by leveraging the effect sizes from the discovery set in a self‐contained test of association in the test dataset.

Interest in the cerebellum is expanding given evidence of its contributions to cognition and emotion, and dysfunction in various psychopathologies. However, research into its genetic architecture and shared influences with liability for mental disorders is lacking. We conducted a genome-wide association study (GWAS) of total cerebellar volume and underlying cerebellar lobe volumes in 33,265 UK-Biobank participants. Total cerebellar volume was heritable (h2SNP = 50.6%), showing moderate genetic homogeneity across lobes (h2SNP from 35.4% to 57.1%; mean genetic correlation between lobes rg ≈ 0.44). We identified 33 GWAS signals associated with total cerebellar volume, of which 6 are known to alter protein-coding gene structure, while a further five mapped to genomic regions known to alter cerebellar tissue gene expression. Use of summary data-based Mendelian randomisation further prioritised genes whose change in expression appears to mediate the SNP-trait association. In total, we highlight 21 unique genes of greatest interest for follow-up analyses. Using LD-regression, we report significant genetic correlations between total cerebellar volume and brainstem, pallidum and thalamus volumes. While the same approach did not result in significant correlations with psychiatric phenotypes, we report enrichment of schizophrenia, bipolar disorder and autism spectrum disorder associated signals within total cerebellar GWAS results via conditional and conjunctional-FDR analysis. Via these methods and GWAS catalogue, we identify which of our cerebellar genomic regions also associate with psychiatric traits. Our results provide important insights into the common allele architecture of cerebellar volume and its overlap with other brain volumes and psychiatric phenotypes.

A major controversy in psychiatric genetics is whether nonadditive genetic interaction effects contribute to the risk of highly polygenic disorders. We applied a support vector machines (SVMs) approach, which is capable of building linear and nonlinear models using kernel methods, to classify cases from controls in a large schizophrenia case–control sample of 11,853 subjects (5,554 cases and 6,299 controls) and compared its prediction accuracy with the polygenic risk score (PRS) approach. We also investigated whether SVMs are a suitable approach to detecting nonlinear genetic effects, that is, interactions. We found that PRS provided more accurate case/control classification than either linear or nonlinear SVMs, and give a tentative explanation why PRS outperforms both multivariate regression and linear kernel SVMs. In addition, we observe that nonlinear kernel SVMs showed higher classification accuracy than linear SVMs when a large number of SNPs are entered into the model. We conclude that SVMs are a potential tool for assessing the presence of interactions, prior to searching for them explicitly.

Late onset Alzheimer’s disease is the most common form of dementia for which about 30 susceptibility loci have been reported. The aim of the current study is to identify novel genes associated with Alzheimer’s disease using the largest up-to-date reference single nucleotide polymorphism (SNP) panel, the most accurate imputation software and a novel gene-based analysis approach which tests for patterns of association within genes, in the powerful genome-wide association dataset of the International Genomics of Alzheimer’s Project Consortium, comprising over 7 million genotypes from 17,008 Alzheimer’s cases and 37,154 controls. In addition to earlier reported genes, we detected three novel gene-wide significant loci PPARGC1A ( p = 2.2 × 10 −6 ), RORA ( p = 7.4 × 10 −7 ) and ZNF423 ( p = 2.1 × 10 −6 ). PPARGC1A and RORA are involved in circadian rhythm; circadian disturbances are one of the earliest symptoms of Alzheimer’s disease. PPARGC1A is additionally linked to energy metabolism and the generation of amyloid beta plaques. RORA is involved in a variety of functions apart from circadian rhythm, such as cholesterol metabolism and inflammation. The ZNF423 gene resides in an Alzheimer’s disease-specific protein network and is likely involved with centrosomes and DNA damage repair.