Macrophages Facilitate Electrical Conduction in the Heart.

Organ-specific functions of tissue-resident macrophages in the steady-state heart are unknown. Here, we show that cardiac macrophages facilitate electrical conduction through the distal atrioventricular node, where conducting cells densely intersperse with elongated macrophages expressing connexin 43. When coupled to spontaneously beating cardiomyocytes via connexin-43-containing gap junctions, cardiac macrophages have a negative resting membrane potential and depolarize in synchrony with cardiomyocytes. Conversely, macrophages render the resting membrane potential of cardiomyocytes more positive and, according to computational modeling, accelerate their repolarization. Photostimulation of channelrhodopsin-2-expressing macrophages improves atrioventricular conduction, whereas conditional deletion of connexin 43 in macrophages and congenital lack of macrophages delay atrioventricular conduction. In the Cd11bDTR mouse, macrophage ablation induces progressive atrioventricular block. These observations implicate macrophages in normal and aberrant cardiac conduction.

Transfer learning enables predictions in network biology.

Mapping gene networks requires large amounts of transcriptomic data to learn the connections between genes, which impedes discoveries in settings with limited data, including rare diseases and diseases affecting clinically inaccessible tissues. Recently, transfer learning has revolutionized fields such as natural language understanding1,2 and computer vision3 by leveraging deep learning models pretrained on large-scale general datasets that can then be fine-tuned towards a vast array of downstream tasks with limited task-specific data. Here, we developed a context-aware, attention-based deep learning model, Geneformer, pretrained on a large-scale corpus of about 30 million single-cell transcriptomes to enable context-specific predictions in settings with limited data in network biology. During pretraining, Geneformer gained a fundamental understanding of network dynamics, encoding network hierarchy in the attention weights of the model in a completely self-supervised manner. Fine-tuning towards a diverse panel of downstream tasks relevant to chromatin and network dynamics using limited task-specific data demonstrated that Geneformer consistently boosted predictive accuracy. Applied to disease modelling with limited patient data, Geneformer identified candidate therapeutic targets for cardiomyopathy. Overall, Geneformer represents a pretrained deep learning model from which fine-tuning towards a broad range of downstream applications can be pursued to accelerate discovery of key network regulators and candidate therapeutic targets.

Single-nucleus profiling of human dilated and hypertrophic cardiomyopathy.

Heart failure encompasses a heterogeneous set of clinical features that converge on impaired cardiac contractile function1,2 and presents a growing public health concern. Previous work has highlighted changes in both transcription and protein expression in failing hearts3,4, but may overlook molecular changes in less prevalent cell types. Here we identify extensive molecular alterations in failing hearts at single-cell resolution by performing single-nucleus RNA sequencing of nearly 600,000 nuclei in left ventricle samples from 11 hearts with dilated cardiomyopathy and 15 hearts with hypertrophic cardiomyopathy as well as 16 non-failing hearts. The transcriptional profiles of dilated or hypertrophic cardiomyopathy hearts broadly converged at the tissue and cell-type level. Further, a subset of hearts from patients with cardiomyopathy harbour a unique population of activated fibroblasts that is almost entirely absent from non-failing samples. We performed a CRISPR-knockout screen in primary human cardiac fibroblasts to evaluate this fibrotic cell state transition; knockout of genes associated with fibroblast transition resulted in a reduction of myofibroblast cell-state transition upon TGFß1 stimulation for a subset of genes. Our results provide insights into the transcriptional diversity of the human heart in health and disease as well as new potential therapeutic targets and biomarkers for heart failure.

Transcriptome variation in human tissues revealed by long-read sequencing.

Regulation of transcript structure generates transcript diversity and plays an important role in human disease1-7. The advent of long-read sequencing technologies offers the opportunity to study the role of genetic variation in transcript structure8-16. In this Article, we present a large human long-read RNA-seq dataset using the Oxford Nanopore Technologies platform from 88 samples from Genotype-Tissue Expression (GTEx) tissues and cell lines, complementing the GTEx resource. We identified just over 70,000 novel transcripts for annotated genes, and validated the protein expression of 10% of novel transcripts. We developed a new computational package, LORALS, to analyse the genetic effects of rare and common variants on the transcriptome by allele-specific analysis of long reads. We characterized allele-specific expression and transcript structure events, providing new insights into the specific transcript alterations caused by common and rare genetic variants and highlighting the resolution gained from long-read data. We were able to perturb the transcript structure upon knockdown of PTBP1, an RNA binding protein that mediates splicing, thereby finding genetic regulatory effects that are modified by the cellular environment. Finally, we used this dataset to enhance variant interpretation and study rare variants leading to aberrant splicing patterns.

Rare variants in long non-coding RNAs are associated with blood lipid levels in the TOPMed whole-genome sequencing study.

Long non-coding RNAs (lncRNAs) are known to perform important regulatory functions in lipid metabolism. Large-scale whole-genome sequencing (WGS) studies and new statistical methods for variant set tests now provide an opportunity to assess more associations between rare variants in lncRNA genes and complex traits across the genome. In this study, we used high-coverage WGS from 66,329 participants of diverse ancestries with measurement of blood lipids and lipoproteins (LDL-C, HDL-C, TC, and TG) in the National Heart, Lung, and Blood Institute (NHLBI) Trans-Omics for Precision Medicine (TOPMed) program to investigate the role of lncRNAs in lipid variability. We aggregated rare variants for 165,375 lncRNA genes based on their genomic locations and conducted rare-variant aggregate association tests using the STAAR (variant-set test for association using annotation information) framework. We performed STAAR conditional analysis adjusting for common variants in known lipid GWAS loci and rare-coding variants in nearby protein-coding genes. Our analyses revealed 83 rare lncRNA variant sets significantly associated with blood lipid levels, all of which were located in known lipid GWAS loci (in a ±500-kb window of a Global Lipids Genetics Consortium index variant). Notably, 61 out of 83 signals (73%) were conditionally independent of common regulatory variation and rare protein-coding variation at the same loci. We replicated 34 out of 61 (56%) conditionally independent associations using the independent UK Biobank WGS data. Our results expand the genetic architecture of blood lipids to rare variants in lncRNAs.

Unsupervised removal of systematic background noise from droplet-based single-cell experiments using CellBender.

Droplet-based single-cell assays, including single-cell RNA sequencing (scRNA-seq), single-nucleus RNA sequencing (snRNA-seq) and cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq), generate considerable background noise counts, the hallmark of which is nonzero counts in cell-free droplets and off-target gene expression in unexpected cell types. Such systematic background noise can lead to batch effects and spurious differential gene expression results. Here we develop a deep generative model based on the phenomenology of noise generation in droplet-based assays. The proposed model accurately distinguishes cell-containing droplets from cell-free droplets, learns the background noise profile and provides noise-free quantification in an end-to-end fashion. We implement this approach in the scalable and robust open-source software package CellBender. Analysis of simulated data demonstrates that CellBender operates near the theoretically optimal denoising limit. Extensive evaluations using real datasets and experimental benchmarks highlight enhanced concordance between droplet-based single-cell data and established gene expression patterns, while the learned background noise profile provides evidence of degraded or uncaptured cell types.

Thrombosis risk in single- and double-heterozygous carriers of factor V Leiden and prothrombin G20210A in FinnGen and the UK Biobank.

ABSTRACT: The factor V Leiden (FVL; rs6025) and prothrombin G20210A (PTGM; rs1799963) polymorphisms are 2 of the most well-studied genetic risk factors for venous thromboembolism (VTE). However, double heterozygosity (DH) for FVL and PTGM remains poorly understood, with previous studies showing marked disagreement regarding thrombosis risk conferred by the DH genotype. Using multidimensional data from the UK Biobank (UKB) and FinnGen biorepositories, we evaluated the clinical impact of DH carrier status across 937 939 individuals. We found that 662 participants (0.07%) were DH carriers. After adjustment for age, sex, and ancestry, DH individuals experienced a markedly elevated risk of VTE compared with wild-type individuals (odds ratio [OR] = 5.24; 95% confidence interval [CI], 4.01-6.84; P = 4.8 × 10-34), which approximated the risk conferred by FVL homozygosity. A secondary analysis restricted to UKB participants (N = 445 144) found that effect size estimates for the DH genotype remained largely unchanged (OR = 4.53; 95% CI, 3.42-5.90; P < 1 × 10-16) after adjustment for commonly cited VTE risk factors, such as body mass index, blood type, and markers of inflammation. In contrast, the DH genotype was not associated with a significantly higher risk of any arterial thrombosis phenotype, including stroke, myocardial infarction, and peripheral artery disease. In summary, we leveraged population-scale genomic data sets to conduct, to our knowledge, the largest study to date on the DH genotype and were able to establish far more precise effect size estimates than previously possible. Our findings indicate that the DH genotype may occur as frequently as FVL homozygosity and may confer a similarly increased risk of VTE.

MMCT-Loop: a mix model-based pipeline for calling targeted 3D chromatin loops.

Protein-specific Chromatin Conformation Capture (3C)-based technologies have become essential for identifying distal genomic interactions with critical roles in gene regulation. The standard techniques include Chromatin Interaction Analysis by Paired-End Tag (ChIA-PET), in situ Hi-C followed by chromatin immunoprecipitation (HiChIP) also known as PLAC-seq. To identify chromatin interactions from these data, a variety of computational methods have emerged. Although these state-of-art methods address many issues with loop calling, only few methods can fit different data types simultaneously, and the accuracy as well as the efficiency these approaches remains limited. Here we have generated a pipeline, MMCT-Loop, which ensures the accurate identification of strong loops as well as dynamic or weak loops through a mixed model. MMCT-Loop outperforms existing methods in accuracy, and the detected loops show higher activation functionality. To highlight the utility of MMCT-Loop, we applied it to conformational data derived from neural stem cell (NSCs) and uncovered several previously unidentified regulatory regions for key master regulators of stem cell identity. MMCT-Loop is an accurate and efficient loop caller for targeted conformation capture data, which supports raw data or pre-processed valid pairs as input, the output interactions are formatted and easily uploaded to a genome browser for visualization.

Clonal Hematopoiesis of Indeterminate Potential With Loss of Tet2 Enhances Risk for Atrial Fibrillation Through Nlrp3 Inflammasome Activation.

BACKGROUND: Clonal hematopoiesis of indeterminate potential (CHIP), a common age-associated phenomenon, associates with increased risk of both hematological malignancy and cardiovascular disease. Although CHIP is known to increase the risk of myocardial infarction and heart failure, the influence of CHIP in cardiac arrhythmias, such as atrial fibrillation (AF), is less explored. METHODS: CHIP prevalence was determined in the UK Biobank, and incident AF analysis was stratified by CHIP status and clone size using Cox proportional hazard models. Lethally irradiated mice were transplanted with hematopoietic-specific loss of Tet2, hematopoietic-specific loss of Tet2 and Nlrp3, or wild-type control and fed a Western diet, compounded with or without NLRP3 (NLR [NACHT, LRR {leucine rich repeat}] family pyrin domain containing protein 3) inhibitor, NP3-361, for 6 to 9 weeks. Mice underwent in vivo invasive electrophysiology studies and ex vivo optical mapping. Cardiomyocytes from Ldlr-/- mice with hematopoietic-specific loss of Tet2 or wild-type control and fed a Western diet were isolated to evaluate calcium signaling dynamics and analysis. Cocultures of pluripotent stem cell-derived atrial cardiomyocytes were incubated with Tet2-deficient bone marrow-derived macrophages, wild-type control, or cytokines IL-1ß (interleukin 1ß) or IL-6 (interleukin 6). RESULTS: Analysis of the UK Biobank showed individuals with CHIP, in particular TET2 CHIP, have increased incident AF. Hematopoietic-specific inactivation of Tet2 increases AF propensity in atherogenic and nonatherogenic mouse models and is associated with increased Nlrp3 expression and CaMKII (Ca2+/calmodulin-dependent protein kinase II) activation, with AF susceptibility prevented by inactivation of Nlrp3. Cardiomyocytes isolated from Ldlr-/- mice with hematopoietic inactivation of Tet2 and fed a Western diet have impaired calcium release from the sarcoplasmic reticulum into the cytosol, contributing to atrial arrhythmogenesis. Abnormal sarcoplasmic reticulum calcium release was recapitulated in cocultures of cardiomyocytes with the addition of Tet2-deficient macrophages or cytokines IL-1ß or IL-6. CONCLUSIONS: We identified a modest association between CHIP, particularly TET2 CHIP, and incident AF in the UK Biobank population. In a mouse model of AF resulting from hematopoietic-specific inactivation of Tet2, we propose altered calcium handling as an arrhythmogenic mechanism, dependent on Nlrp3 inflammasome activation. Our data are in keeping with previous studies of CHIP in cardiovascular disease, and further studies into the therapeutic potential of NLRP3 inhibition for individuals with TET2 CHIP may be warranted.

Rare coding variants in 35 genes associate with circulating lipid levels-A multi-ancestry analysis of 170,000 exomes.

Large-scale gene sequencing studies for complex traits have the potential to identify causal genes with therapeutic implications. We performed gene-based association testing of blood lipid levels with rare (minor allele frequency < 1%) predicted damaging coding variation by using sequence data from >170,000 individuals from multiple ancestries: 97,493 European, 30,025 South Asian, 16,507 African, 16,440 Hispanic/Latino, 10,420 East Asian, and 1,182 Samoan. We identified 35 genes associated with circulating lipid levels; some of these genes have not been previously associated with lipid levels when using rare coding variation from population-based samples. We prioritize 32 genes in array-based genome-wide association study (GWAS) loci based on aggregations of rare coding variants; three (EVI5, SH2B3, and PLIN1) had no prior association of rare coding variants with lipid levels. Most of our associated genes showed evidence of association among multiple ancestries. Finally, we observed an enrichment of gene-based associations for low-density lipoprotein cholesterol drug target genes and for genes closest to GWAS index single-nucleotide polymorphisms (SNPs). Our results demonstrate that gene-based associations can be beneficial for drug target development and provide evidence that the gene closest to the array-based GWAS index SNP is often the functional gene for blood lipid levels.

Loss of the Atrial Fibrillation-Related Gene, Zfhx3, Results in Atrial Dilation and Arrhythmias.

BACKGROUND: ZFHX3 (zinc finger homeobox 3), a gene that encodes a large transcription factor, is at the second-most significantly associated locus with atrial fibrillation (AF), but its function in the heart is unknown. This study aims to identify causative genetic variation related to AF at the ZFHX3 locus and examine the impact of Zfhx3 loss on cardiac function in mice. METHODS: CRISPR-Cas9 genome editing, chromatin immunoprecipitation, and luciferase assays in pluripotent stem cell-derived cardiomyocytes were used to identify causative genetic variation related to AF at the ZFHX3 locus. Cardiac function was assessed by echocardiography, magnetic resonance imaging, electrophysiology studies, calcium imaging, and RNA sequencing in mice with heterozygous and homozygous cardiomyocyte-restricted Zfhx3 loss (Zfhx3 Het and knockout, respectively). Human cardiac single-nucleus ATAC (assay for transposase-accessible chromatin)-sequencing data was analyzed to determine which genes in atrial cardiomyocytes are directly regulated by ZFHX3. RESULTS: We found single-nucleotide polymorphism (SNP) rs12931021 modulates an enhancer regulating ZFHX3 expression, and the AF risk allele is associated with decreased ZFHX3 transcription. We observed a gene-dose response in AF susceptibility with Zfhx3 knockout mice having higher incidence, frequency, and burden of AF than Zfhx3 Het and wild-type mice, with alterations in conduction velocity, atrial action potential duration, calcium handling and the development of atrial enlargement and thrombus, and dilated cardiomyopathy. Zfhx3 loss results in atrial-specific differential effects on genes and signaling pathways involved in cardiac pathophysiology and AF. CONCLUSIONS: Our findings implicate ZFHX3 as the causative gene at the 16q22 locus for AF, and cardiac abnormalities caused by loss of cardiac Zfhx3 are due to atrial-specific dysregulation of pathways involved in AF susceptibility. Together, these data reveal a novel and important role for Zfhx3 in the control of cardiac genes and signaling pathways essential for normal atrial function.

Using Genomics to Identify Novel Therapeutic Targets for Aortic Disease.

Aortic disease, including dissection, aneurysm, and rupture, carries significant morbidity and mortality and is a notable cause of sudden cardiac death. Much of our knowledge regarding the genetic basis of aortic disease has relied on the study of individuals with Mendelian aortopathies and, until recently, the genetic determinants of population-level variance in aortic phenotypes remained unclear. However, the application of machine learning methodologies to large imaging datasets has enabled researchers to rapidly define aortic traits and mine dozens of novel genetic associations for phenotypes such as aortic diameter and distensibility. In this review, we highlight the emerging potential of genomics for identifying causal genes and candidate drug targets for aortic disease. We describe how deep learning technologies have accelerated the pace of genetic discovery in this field. We then provide a blueprint for translating genetic associations to biological insights, reviewing techniques for locus and cell type prioritization, high-throughput functional screening, and disease modeling using cellular and animal models of aortic disease.

Clonal haematopoiesis of indeterminate potential predicts incident cardiac arrhythmias.

BACKGROUND AND AIMS: Clonal haematopoiesis of indeterminate potential (CHIP), the age-related expansion of blood cells with preleukemic mutations, is associated with atherosclerotic cardiovascular disease and heart failure. This study aimed to test the association of CHIP with new-onset arrhythmias. METHODS: UK Biobank participants without prevalent arrhythmias were included. Co-primary study outcomes were supraventricular arrhythmias, bradyarrhythmias, and ventricular arrhythmias. Secondary outcomes were cardiac arrest, atrial fibrillation, and any arrhythmia. Associations of any CHIP [variant allele fraction (VAF) ≥ 2%], large CHIP (VAF ≥10%), and gene-specific CHIP subtypes with incident arrhythmias were evaluated using multivariable-adjusted Cox regression. Associations of CHIP with myocardial interstitial fibrosis [T1 measured using cardiac magnetic resonance (CMR)] were also tested. RESULTS: This study included 410 702 participants [CHIP: n = 13 892 (3.4%); large CHIP: n = 9191 (2.2%)]. Any and large CHIP were associated with multi-variable-adjusted hazard ratios of 1.11 [95% confidence interval (CI) 1.04-1.18; P = .001] and 1.13 (95% CI 1.05-1.22; P = .001) for supraventricular arrhythmias, 1.09 (95% CI 1.01-1.19; P = .031) and 1.13 (95% CI 1.03-1.25; P = .011) for bradyarrhythmias, and 1.16 (95% CI, 1.00-1.34; P = .049) and 1.22 (95% CI 1.03-1.45; P = .021) for ventricular arrhythmias, respectively. Associations were independent of coronary artery disease and heart failure. Associations were also heterogeneous across arrhythmia subtypes and strongest for cardiac arrest. Gene-specific analyses revealed an increased risk of arrhythmias across driver genes other than DNMT3A. Large CHIP was associated with 1.31-fold odds (95% CI 1.07-1.59; P = .009) of being in the top quintile of myocardial fibrosis by CMR. CONCLUSIONS: CHIP may represent a novel risk factor for incident arrhythmias, indicating a potential target for modulation towards arrhythmia prevention and treatment.

TAILS Identifies Candidate Substrates and Biomarkers of ADAMTS7, a Therapeutic Protease Target in Coronary Artery Disease.

Loss-of-function mutations in the secreted enzyme ADAMTS7 (a disintegrin and metalloproteinase with thrombospondin motifs 7) are associated with protection for coronary artery disease. ADAMTS7 catalytic inhibition has been proposed as a therapeutic strategy for treating coronary artery disease; however, the lack of an endogenous substrate has hindered the development of activity-based biomarkers. To identify ADAMTS7 extracellular substrates and their cleavage sites relevant to vascular disease, we used TAILS (terminal amine isotopic labeling of substrates), a method for identifying protease-generated neo-N termini. We compared the secreted proteome of vascular smooth muscle and endothelial cells expressing either full-length mouse ADAMTS7 WT, catalytic mutant ADAMTS7 E373Q, or a control luciferase adenovirus. Significantly enriched N-terminal cleavage sites in ADAMTS7 WT samples were compared to the negative control conditions and filtered for stringency, resulting in catalogs of high confidence candidate ADAMTS7 cleavage sites from our three independent TAILS experiments. Within the overlap of these discovery sets, we identified 24 unique cleavage sites from 16 protein substrates, including cleavage sites in EFEMP1 (EGF-containing fibulin-like extracellular matrix protein 1/Fibulin-3). The ADAMTS7 TAILS preference for EFEMP1 cleavage at the amino acids 123.124 over the adjacent 124.125 site was validated using both endogenous EFEMP1 and purified EFEMP1 in a binary in vitro cleavage assay. Collectively, our TAILS discovery experiments have uncovered hundreds of potential substrates and cleavage sites to explore disease-related biological substrates and facilitate activity-based ADAMTS7 biomarker development.

A polygenic risk score predicts atrial fibrillation in cardiovascular disease.

AIMS: Interest in targeted screening programmes for atrial fibrillation (AF) has increased, yet the role of genetics in identifying patients at highest risk of developing AF is unclear. METHODS AND RESULTS: A total of 36,662 subjects without prior AF were analyzed from four TIMI trials. Subjects were divided into quintiles using a validated polygenic risk score (PRS) for AF. Clinical risk for AF was calculated using the CHARGE-AF model. Kaplan-Meier event rates, adjusted hazard ratios (HRs), C-indices, and net reclassification improvement were used to determine if the addition of the PRS improved prediction compared with clinical risk and N-terminal pro-B-type natriuretic peptide (NT-proBNP). Over 2.3 years, 1018 new AF cases developed. AF PRS predicted a significant risk gradient for AF with a 40% increased risk per 1-SD increase in PRS [HR: 1.40 (1.32-1.49); P < 0.001]. Those with high AF PRS (top 20%) were more than two-fold more likely to develop AF [HR 2.45 (1.99-3.03), P < 0.001] compared with low PRS (bottom 20%). Furthermore, PRS provided an additional gradient of risk stratification on top of the CHARGE-AF clinical risk score, ranging from a 3-year incidence of 1.3% in patients with low clinical and genetic risk to 8.7% in patients with high clinical and genetic risk. The subgroup of patients with high clinical risk, high PRS, and elevated NT-proBNP had an AF risk of 16.7% over 3 years. The C-index with the CHARGE-AF clinical risk score alone was 0.65, which improved to 0.67 (P < 0.001) with the addition of NT-proBNP, and increased further to 0.70 (P < 0.001) with the addition of the PRS. CONCLUSION: In patients with cardiovascular conditions, AF PRS is a strong independent predictor of incident AF that provides complementary predictive value when added to a validated clinical risk score and NT-proBNP.

Screening for Atrial Fibrillation in Older Adults at Primary Care Visits: VITAL-AF Randomized Controlled Trial.

BACKGROUND: Undiagnosed atrial fibrillation (AF) may cause preventable strokes. Guidelines differ regarding AF screening recommendations. We tested whether point-of-care screening with a handheld single-lead ECG at primary care practice visits increases diagnoses of AF. METHODS: We randomized 16 primary care clinics 1:1 to AF screening using a handheld single-lead ECG (AliveCor KardiaMobile) during vital sign assessments, or usual care. Patients included were ages ≥65 years. Screening results were provided to primary care clinicians at the encounter. All confirmatory diagnostic testing and treatment decisions were made by the primary care clinician. New AF diagnoses during the 1-year follow-up were ascertained electronically and manually adjudicated. Proportions and incidence rates were calculated. Effect heterogeneity was assessed. RESULTS: Of 30 715 patients without prevalent AF (n=15 393 screening [91% screened], n=15 322 control), 1.72% of individuals in the screening group had new AF diagnosed at 1 year versus 1.59% in the control group (risk difference, 0.13% [95% CI, -0.16 to 0.42]; P=0.38). In prespecified subgroup analyses, new AF diagnoses in the screening and control groups were greater among those aged ≥85 years (5.56% versus 3.76%, respectively; risk difference, 1.80% [95% CI, 0.18 to 3.30]). The difference in newly diagnosed AF between the screening period and the previous year was marginally greater in the screening versus control group (0.32% versus -0.12%; risk difference, 0.43% [95% CI, -0.01 to 0.84]). The proportion of individuals with newly diagnosed AF who were initiated on oral anticoagulants was not different in the screening (n=194, 73.5%) and control (n=172, 70.8%) arms (risk difference, 2.7% [95% CI, -5.5 to 10.4]). CONCLUSIONS: Screening for AF using a single-lead ECG at primary care visits did not affect new AF diagnoses among all individuals aged 65 years or older compared with usual care. REGISTRATION: URL: https://www. CLINICALTRIALS: gov; Unique identifier: NCT03515057.

Deep Learning of the Retina Enables Phenome- and Genome-Wide Analyses of the Microvasculature.

BACKGROUND: The microvasculature, the smallest blood vessels in the body, has key roles in maintenance of organ health and tumorigenesis. The retinal fundus is a window for human in vivo noninvasive assessment of the microvasculature. Large-scale complementary machine learning-based assessment of the retinal vasculature with phenome-wide and genome-wide analyses may yield new insights into human health and disease. METHODS: We used 97 895 retinal fundus images from 54 813 UK Biobank participants. Using convolutional neural networks to segment the retinal microvasculature, we calculated vascular density and fractal dimension as a measure of vascular branching complexity. We associated these indices with 1866 incident International Classification of Diseases-based conditions (median 10-year follow-up) and 88 quantitative traits, adjusting for age, sex, smoking status, and ethnicity. RESULTS: Low retinal vascular fractal dimension and density were significantly associated with higher risks for incident mortality, hypertension, congestive heart failure, renal failure, type 2 diabetes, sleep apnea, anemia, and multiple ocular conditions, as well as corresponding quantitative traits. Genome-wide association of vascular fractal dimension and density identified 7 and 13 novel loci, respectively, that were enriched for pathways linked to angiogenesis (eg, vascular endothelial growth factor, platelet-derived growth factor receptor, angiopoietin, and WNT signaling pathways) and inflammation (eg, interleukin, cytokine signaling). CONCLUSIONS: Our results indicate that the retinal vasculature may serve as a biomarker for future cardiometabolic and ocular disease and provide insights into genes and biological pathways influencing microvascular indices. Moreover, such a framework highlights how deep learning of images can quantify an interpretable phenotype for integration with electronic health record, biomarker, and genetic data to inform risk prediction and risk modification.

ECG-Based Deep Learning and Clinical Risk Factors to Predict Atrial Fibrillation.

BACKGROUND: Artificial intelligence (AI)-enabled analysis of 12-lead ECGs may facilitate efficient estimation of incident atrial fibrillation (AF) risk. However, it remains unclear whether AI provides meaningful and generalizable improvement in predictive accuracy beyond clinical risk factors for AF. METHODS: We trained a convolutional neural network (ECG-AI) to infer 5-year incident AF risk using 12-lead ECGs in patients receiving longitudinal primary care at Massachusetts General Hospital (MGH). We then fit 3 Cox proportional hazards models, composed of ECG-AI 5-year AF probability, CHARGE-AF clinical risk score (Cohorts for Heart and Aging in Genomic Epidemiology-Atrial Fibrillation), and terms for both ECG-AI and CHARGE-AF (CH-AI), respectively. We assessed model performance by calculating discrimination (area under the receiver operating characteristic curve) and calibration in an internal test set and 2 external test sets (Brigham and Women's Hospital [BWH] and UK Biobank). Models were recalibrated to estimate 2-year AF risk in the UK Biobank given limited available follow-up. We used saliency mapping to identify ECG features most influential on ECG-AI risk predictions and assessed correlation between ECG-AI and CHARGE-AF linear predictors. RESULTS: The training set comprised 45 770 individuals (age 55±17 years, 53% women, 2171 AF events) and the test sets comprised 83 162 individuals (age 59±13 years, 56% women, 2424 AF events). Area under the receiver operating characteristic curve was comparable using CHARGE-AF (MGH, 0.802 [95% CI, 0.767-0.836]; BWH, 0.752 [95% CI, 0.741-0.763]; UK Biobank, 0.732 [95% CI, 0.704-0.759]) and ECG-AI (MGH, 0.823 [95% CI, 0.790-0.856]; BWH, 0.747 [95% CI, 0.736-0.759]; UK Biobank, 0.705 [95% CI, 0.673-0.737]). Area under the receiver operating characteristic curve was highest using CH-AI (MGH, 0.838 [95% CI, 0.807 to 0.869]; BWH, 0.777 [95% CI, 0.766 to 0.788]; UK Biobank, 0.746 [95% CI, 0.716 to 0.776]). Calibration error was low using ECG-AI (MGH, 0.0212; BWH, 0.0129; UK Biobank, 0.0035) and CH-AI (MGH, 0.012; BWH, 0.0108; UK Biobank, 0.0001). In saliency analyses, the ECG P-wave had the greatest influence on AI model predictions. ECG-AI and CHARGE-AF linear predictors were correlated (Pearson r: MGH, 0.61; BWH, 0.66; UK Biobank, 0.41). CONCLUSIONS: AI-based analysis of 12-lead ECGs has similar predictive usefulness to a clinical risk factor model for incident AF and the approaches are complementary. ECG-AI may enable efficient quantification of future AF risk.

Monogenic and Polygenic Contributions to QTc Prolongation in the Population.

BACKGROUND: Rare sequence variation in genes underlying cardiac repolarization and common polygenic variation influence QT interval duration. However, current clinical genetic testing of individuals with unexplained QT prolongation is restricted to examination of monogenic rare variants. The recent emergence of large-scale biorepositories with sequence data enables examination of the joint contribution of rare and common variations to the QT interval in the population. METHODS: We performed a genome-wide association study of the QTc in 84 630 UK Biobank participants and created a polygenic risk score (PRS). Among 26 976 participants with whole-genome sequencing and ECG data in the TOPMed (Trans-Omics for Precision Medicine) program, we identified 160 carriers of putative pathogenic rare variants in 10 genes known to be associated with the QT interval. We examined QTc associations with the PRS and with rare variants in TOPMed. RESULTS: Fifty-four independent loci were identified by genome-wide association study in the UK Biobank. Twenty-one loci were novel, of which 12 were replicated in TOPMed. The PRS composed of 1 110 494 common variants was significantly associated with the QTc in TOPMed (ΔQTc/decile of PRS=1.4 ms [95% CI, 1.3 to 1.5]; P=1.1×10-196). Carriers of putative pathogenic rare variants had longer QTc than noncarriers (ΔQTc=10.9 ms [95% CI, 7.4 to 14.4]). Of individuals with QTc>480 ms, 23.7% carried either a monogenic rare variant or had a PRS in the top decile (3.4% monogenic, 21% top decile of PRS). CONCLUSIONS: QTc duration in the population is influenced by both rare variants in genes underlying cardiac repolarization and polygenic risk, with a sizeable contribution from polygenic risk. Comprehensive assessment of the genetic determinants of QTc prolongation includes incorporation of both polygenic and monogenic risk.

Clinical and Genetic Atrial Fibrillation Risk and Discrimination of Cardioembolic From Noncardioembolic Stroke.

BACKGROUND: Stroke is a leading cause of death and disability worldwide. Atrial fibrillation (AF) is a common cause of stroke but may not be detectable at the time of stroke. We hypothesized that an AF polygenic risk score (PRS) can discriminate between cardioembolic stroke and noncardioembolic strokes. METHODS: We evaluated AF and stroke risk in 26 145 individuals of European descent from the Stroke Genetics Network case-control study. AF genetic risk was estimated using 3 recently developed PRS methods (LDpred-funct-inf, sBayesR, and PRS-CS) and 2 previously validated PRSs. We performed logistic regression of each AF PRS on AF status and separately cardioembolic stroke, adjusting for clinical risk score (CRS), imputation group, and principal components. We calculated model discrimination of AF and cardioembolic stroke using the concordance statistic (c-statistic) and compared c-statistics using 2000-iteration bootstrapping. We also assessed reclassification of cardioembolic stroke with the addition of PRS to either CRS or a modified CHA2DS2-VASc score alone. RESULTS: Each AF PRS was significantly associated with AF and with cardioembolic stroke after adjustment for CRS. Addition of each AF PRS significantly improved discrimination as compared with CRS alone (P<0.01). When combined with the CRS, both PRS-CS and LDpred scores discriminated both AF and cardioembolic stroke (c-statistic 0.84 for AF; 0.74 for cardioembolic stroke) better than 3 other PRS scores (P<0.01). Using PRS-CS PRS and CRS in combination resulted in more appropriate reclassification of stroke events as compared with CRS alone (event reclassification [net reclassification indices]+=14% [95% CI, 10%-18%]; nonevent reclassification [net reclassification indices]-=17% [95% CI, 15%-0.19%]) or the modified CHA2DS2-VASc score (net reclassification indices+=11% [95% CI, 7%-15%]; net reclassification indices-=14% [95% CI, 12%-16%]) alone. CONCLUSIONS: Addition of polygenic risk of AF to clinical risk factors modestly improves the discrimination of cardioembolic from noncardioembolic strokes, as well as reclassification of stroke subtype. Polygenic risk of AF may be a useful biomarker for identifying strokes caused by AF.