Improving multi-population genomic prediction accuracy using multi-trait GBLUP models which incorporate global or local genetic correlation information.

In the application of genomic prediction, a situation often faced is that there are multiple populations in which genomic prediction (GP) need to be conducted. A common way to handle the multi-population GP is simply to combine the multiple populations into a single population. However, since these populations may be subject to different environments, there may exist genotype-environment interactions which may affect the accuracy of genomic prediction. In this study, we demonstrated that multi-trait genomic best linear unbiased prediction (MTGBLUP) can be used for multi-population genomic prediction, whereby the performances of a trait in different populations are regarded as different traits, and thus multi-population prediction is regarded as multi-trait prediction by employing the between-population genetic correlation. Using real datasets, we proved that MTGBLUP outperformed the conventional multi-population model that simply combines different populations together. We further proposed that MTGBLUP can be improved by partitioning the global between-population genetic correlation into local genetic correlations (LGC). We suggested two LGC models, LGC-model-1 and LGC-model-2, which partition the genome into regions with and without significant LGC (LGC-model-1) or regions with and without strong LGC (LGC-model-2). In analysis of real datasets, we demonstrated that the LGC models could increase universally the prediction accuracy and the relative improvement over MTGBLUP reached up to 163.86% (25.64% on average).

Benchmarking of local genetic correlation estimation methods using summary statistics from genome-wide association studies.

Local genetic correlation evaluates the correlation of additive genetic effects between different traits across the same genetic variants at a genomic locus. It has been proven informative for understanding the genetic similarities of complex traits beyond that captured by global genetic correlation calculated across the whole genome. Several summary-statistics-based approaches have been developed for estimating local genetic correlation, including $\rho$-hess, SUPERGNOVA and LAVA. However, there has not been a comprehensive evaluation of these methods to offer practical guidelines on the choices of these methods. In this study, we conduct benchmark comparisons of the performance of these three methods through extensive simulation and real data analyses. We focus on two technical difficulties in estimating local genetic correlation: sample overlaps across traits and local linkage disequilibrium (LD) estimates when only the external reference panels are available. Our simulations suggest the likelihood of incorrectly identifying correlated regions and local correlation estimation accuracy are highly dependent on the estimation of the local LD matrix. These observations are corroborated by real data analyses of 31 complex traits. Overall, our findings illuminate the distinct results yielded by different methods applied in post-genome-wide association studies (post-GWAS) local correlation studies. We underscore the sensitivity of local genetic correlation estimates and inferences to the precision of local LD estimation. These observations accentuate the vital need for ongoing refinement in methodologies.

Dissecting loci that underpin the genetic correlations between production, fertility, and urea traits in Australian Holstein cattle.

Unfavorable genetic correlations between milk production, fertility, and urea traits have been reported. However, knowledge of the genomic regions associated with these unfavorable correlations is limited. Here, we used the correlation scan method to identify and investigate the regions driving or antagonizing the genetic correlations between production vs. fertility, urea vs. fertility, and urea vs. production traits. Driving regions produce an estimate of correlation that is in the same direction as the global correlation. Antagonizing regions produce an estimate in the opposite direction of the global estimates. Our dataset comprised 6567, 4700, and 12,658 Holstein cattle with records of production traits (milk yield, fat yield, and protein yield), fertility (calving interval) and urea traits (milk urea nitrogen and blood urea nitrogen predicted using milk-mid-infrared spectroscopy), respectively. Several regions across the genome drive the correlations between production, fertility, and urea traits. Antagonizing regions were confined to certain parts of the genome and the genes within these regions were mostly involved in preventing metabolic dysregulation, liver reprogramming, metabolism remodeling, and lipid homeostasis. The driving regions were enriched for QTL related to puberty, milk, and health-related traits. Antagonizing regions were mostly related to muscle development, metabolic body weight, and milk traits. In conclusion, we have identified genomic regions of potential importance for dairy cattle breeding. Future studies could investigate the antagonizing regions as potential genomic regions to break the unfavorable correlations and improve milk production as well as fertility and urea traits.

Investigating Genetic Overlap between Alzheimer's Disease, Lipids, and Coronary Artery Disease: A Large-Scale Genome-Wide Cross Trait Analysis.

There is evidence to support a link between abnormal lipid metabolism and Alzheimer's disease (AD) risk. Similarly, observational studies suggest a comorbid relationship between AD and coronary artery disease (CAD). However, the intricate biological mechanisms of AD are poorly understood, and its relationship with lipids and CAD traits remains unresolved. Conflicting evidence further underscores the ongoing investigation into this research area. Here, we systematically assess the cross-trait genetic overlap of AD with 13 representative lipids (from eight classes) and seven CAD traits, leveraging robust analytical methods, well-powered large-scale genetic data, and rigorous replication testing. Our main analysis demonstrates a significant positive global genetic correlation of AD with triglycerides and all seven CAD traits assessed-angina pectoris, cardiac dysrhythmias, coronary arteriosclerosis, ischemic heart disease, myocardial infarction, non-specific chest pain, and coronary artery disease. Gene-level analyses largely reinforce these findings and highlight the genetic overlap between AD and three additional lipids: high-density lipoproteins (HDLs), low-density lipoproteins (LDLs), and total cholesterol. Moreover, we identify genome-wide significant genes (Fisher's combined p value [FCPgene] < 2.60 × 10-6) shared across AD, several lipids, and CAD traits, including WDR12, BAG6, HLA-DRA, PHB, ZNF652, APOE, APOC4, PVRL2, and TOMM40. Mendelian randomisation analysis found no evidence of a significant causal relationship between AD, lipids, and CAD traits. However, local genetic correlation analysis identifies several local pleiotropic hotspots contributing to the relationship of AD with lipids and CAD traits across chromosomes 6, 8, 17, and 19. Completing a three-way analysis, we confirm a strong genetic correlation between lipids and CAD traits-HDL and sphingomyelin demonstrate negative correlations, while LDL, triglycerides, and total cholesterol show positive correlations. These findings support genetic overlap between AD, specific lipids, and CAD traits, implicating shared but non-causal genetic susceptibility. The identified shared genes and pleiotropic hotspots are valuable targets for further investigation into AD and, potentially, its comorbidity with CAD traits.

Genetic Architecture of Ischaemic Strokes after COVID-19 Shows Similarities with Large Vessel Strokes.

We aimed to analyse whether patients with ischaemic stroke (IS) occurring within eight days after the onset of COVID-19 (IS-COV) are associated with a specific aetiology of IS. We used SUPERGNOVA to identify genome regions that correlate between the IS-COV cohort (73 IS-COV cases vs. 701 population controls) and different aetiological subtypes. Polygenic risk scores (PRSs) for each subtype were generated and tested in the IS-COV cohort using PRSice-2 and PLINK to find genetic associations. Both analyses used the IS-COV cohort and GWAS from MEGASTROKE (67,162 stroke patients vs. 454,450 population controls), GIGASTROKE (110,182 vs. 1,503,898), and the NINDS Stroke Genetics Network (16,851 vs. 32,473). Three genomic regions were associated (p-value < 0.05) with large artery atherosclerosis (LAA) and cardioembolic stroke (CES). We found four loci targeting the genes PITX2 (rs10033464, IS-COV beta = 0.04, p-value = 2.3 × 10-2, se = 0.02), previously associated with CES, HS6ST1 (rs4662630, IS-COV beta = -0.04, p-value = 1.3 × 10-3, se = 0.01), TMEM132E (rs12941838 IS-COV beta = 0.05, p-value = 3.6 × 10-4, se = 0.01), and RFFL (rs797989 IS-COV beta = 0.03, p-value = 1.0 × 10-2, se = 0.01). A statistically significant PRS was observed for LAA. Our results suggest that IS-COV cases are genetically similar to LAA and CES subtypes. Larger cohorts are needed to assess if the genetic factors in IS-COV cases are shared with the general population or specific to viral infection.

Relationship of Cognition and Alzheimer's Disease with Gastrointestinal Tract Disorders: A Large-Scale Genetic Overlap and Mendelian Randomisation Analysis.

Emerging observational evidence suggests links between cognitive impairment and a range of gastrointestinal tract (GIT) disorders; however, the mechanisms underlying their relationships remain unclear. Leveraging large-scale genome-wide association studies' summary statistics, we comprehensively assessed genetic overlap and potential causality of cognitive traits and Alzheimer's disease (AD) with several GIT disorders. We demonstrate a strong and highly significant inverse global genetic correlation between cognitive traits and GIT disorders­peptic ulcer disease (PUD), gastritis-duodenitis, diverticulosis, irritable bowel syndrome, and gastroesophageal reflux disease (GERD), but not inflammatory bowel disease (IBD). Further analysis detects 35 significant (p < 4.37 × 10−5) bivariate local genetic correlations between cognitive traits, AD, and GIT disorders (including IBD). Mendelian randomisation analysis suggests a risk-decreasing causality of educational attainment, intelligence, and other cognitive traits on PUD and GERD, but not IBD, and a putative association of GERD with cognitive function decline. Gene-based analysis reveals a significant gene-level genetic overlap of cognitive traits with AD and GIT disorders (IBD inclusive, pbinomial-test = 1.18 × 10−3−2.20 × 10−16). Our study supports the protective roles of genetically-influenced educational attainments and other cognitive traits on the risk of GIT disorders and highlights a putative association of GERD with cognitive function decline. Findings from local genetic correlation analysis provide novel insights, indicating that the relationship of IBD with cognitive traits (and AD) will depend largely on their local effects across the genome.

Local patterns of genetic sharing challenge the boundaries between neuropsychiatric and insulin resistance-related conditions.

The co-occurrence of insulin resistance (IR)-related metabolic conditions with neuropsychiatric disorders is a complex public health challenge. Evidence of the genetic links between these phenotypes is emerging, but little is currently known about the genomic regions and biological functions that are involved. To address this, we performed Local Analysis of [co]Variant Association (LAVA) using large-scale (N=9,725-933,970) genome-wide association studies (GWASs) results for three IR-related conditions (type 2 diabetes mellitus, obesity, and metabolic syndrome) and nine neuropsychiatric disorders. Subsequently, positional and expression quantitative trait locus (eQTL)-based gene mapping and downstream functional genomic analyses were performed on the significant loci. Patterns of negative and positive local genetic correlations (|rg|=0.21-1, pFDR<0.05) were identified at 109 unique genomic regions across all phenotype pairs. Local correlations emerged even in the absence of global genetic correlations between IR-related conditions and Alzheimer's disease, bipolar disorder, and Tourette's syndrome. Genes mapped to the correlated regions showed enrichment in biological pathways integral to immune-inflammatory function, vesicle trafficking, insulin signalling, oxygen transport, and lipid metabolism. Colocalisation analyses further prioritised 10 genetically correlated regions for likely harbouring shared causal variants, displaying high deleterious or regulatory potential. These variants were found within or in close proximity to genes, such as SLC39A8 and HLA-DRB1, that can be targeted by supplements and already known drugs, including omega-3/6 fatty acids, immunomodulatory, antihypertensive, and cholesterol-lowering drugs. Overall, our findings underscore the complex genetic landscape of IR-neuropsychiatric multimorbidity, advocating for an integrated disease model and offering novel insights for research and treatment strategies in this domain.

The genetic overlap between Alzheimer's disease, amyotrophic lateral sclerosis, Lewy body dementia, and Parkinson's disease.

Neurodegenerative diseases are a group of disorders characterized by neuronal cell death causing a variety of physical and mental problems. While these disorders can be characterized by their phenotypic presentation within the nervous system, their aetiologies differ to varying degrees. The majority of previous genetic evidence for overlap between neurodegenerative diseases has been pairwise. In this study, we aimed to identify overlap between the 4 investigated neurodegenerative disorders (Alzheimer's disease, amyotrophic lateral sclerosis, Lewy body dementia, and Parkinson's disease) at the variant, gene, genomic locus, gene-set, cell, or tissue level, with specific interest in overlap between 3 or more diseases. Using local genetic correlation, we found 2 loci (TMEM175 and HLA) that were shared across 3 disorders. We also highlighted genes, genomic loci, gene sets, cell types, and tissue types which may be important to 2 or more disorders by analyzing the association of variants with a common factor estimated from the 4 disorders. Our study successfully highlighted genetic loci and tissues associated with 2 or more neurodegenerative diseases.

A Local Genetic Correlation Analysis Provides Biological Insights Into the Shared Genetic Architecture of Psychiatric and Substance Use Phenotypes.

BACKGROUND: Global genetic correlation analysis has provided valuable insight into the shared genetic basis between psychiatric and substance use disorders. However, little is known about which regions disproportionately contribute to the global correlation. METHODS: We used Local Analysis of [co]Variant Annotation to calculate bivariate local genetic correlations across 2495 approximately equal-sized, semi-independent genomic regions for 20 psychiatric and substance use phenotypes. We performed a transcriptome-wide association study using expression weights from the prefrontal cortex to identify risk genes for each phenotype, followed by probabilistic fine-mapping to prioritize credible causal genes within each bivariate locus. RESULTS: We detected 80 significant (p < 2.08 × 10-6) bivariate local genetic correlations across 61 loci. The expression effect directions for risk genes within each bivariate locus were largely consistent with the local correlation coefficients, suggesting that genetically regulated gene expression may be used in the functional interpretation of local genetic correlations. Probabilistic fine-mapping identified several genes that may drive pleiotropic mechanisms for genetically correlated phenotypes. For example, we confirmed a local genetic correlation between schizophrenia and smoking behavior at 15q25 and prioritized PSMA4 as the most credible gene candidate underlying both phenotypes. CONCLUSIONS: Our study reveals previously unreported local bivariate genetic correlations between psychiatric and substance use phenotypes, which we fine-mapped to identify shared credible causal genes underlying genetically correlated phenotypes.