Genetics in Medicine

Background: Many at-risk patients lack access to genetic services due to a genetic counselor (GC) workforce shortage. Little is known about how digital alternatives impact patients with and without cancer who meet criteria for genetic testing. Methods: eREACH2 is a randomized 4-arm non-inferiority trial where pre-test (visit 1) and/or return of results (visit 2) GC counseling was replaced with a patient-centered digital intervention. Arms include: A (GC/GC), B (GC/digital), C (digital/GC) and D (digital/digital). Primary outcomes were non-inferiority in uptake of genetic services and change in genetic knowledge and general anxiety from baseline to post-disclosure of results (T0-T2). Secondary cognitive and affective outcomes were assessed using non-inferiority ANOVAs and equivalency chi-squared tests in intention-to-treat and per-protocol analyses. Findings: 773 participants were recruited nationwide; 46.6% from rural areas. Mean age was 51 years (range 20-87), 13% male, 12% non-white, 29% had less than a college education, and 33% had a personal history of cancer. 584 (76%) patients completed testing (14% had a positive result, 16% had a VUS). In the primary ITT analyses, we met the non-inferiority for uptake of genetic services and anxiety, but results were inconclusive for knowledge. Secondary outcomes were heterogeneous across arms. Arm C demonstrated consistently favorable effects, while Arms B and D showed less favorable outcomes in select domains (e.g. satisfaction and MICRA). Patients who received positive or VUS results via digital disclosure had significantly higher MICRA scores - indicating greater negative response to testing. Interpretation: In this large, randomized trial of patients with and without cancer, the eREACH intervention was effective for pre-test counseling, but inconclusive for digital disclosure of results. Exploratory analyses suggest that digital delivery could be a reasonable alternative for individuals receiving negative results, while those receiving positive or VUS results may derive some short-term psychosocial benefit from GC disclosure.

Sub-Saharan African populations carry pharmacogenomic alleles poorly represented in the European-derived reference panels underlying most clinical genotyping tools. We present a curated, machine-readable catalog of nine actionable alleles across six pharmacogenes (CYP2D6, CYP2B6, CYP2C9, CYP2C19, CYP3A5, NAT2) with African-specific frequency ranges, functional annotations, and evidence levels derived from reanalysis of 661 high-coverage whole-genome sequences across seven 1000 Genomes Project African populations. Direct comparison against PharmCAT v3.4.0 shows that CYP2D6 produces zero diplotype calls (0/661 samples callable) due to monomorphic reference positions absent from standard variant-only VCF output, a known limitation whose consequences for African allele carriers had not been reported. afripharmagen's reduced-position strategy identifies 243 CYP2D617 and 134 CYP2D629 carriers from the same input. For CYP2B6, CYP2C9, CYP2C19, and NAT2, both tools show concordance of 95-100%. Frequency gradients (CYP2B66: 30-50%; CYP2D617: 15-35% in West Africa; CYP3A5*1: 60-95%) translate directly into prescribing risk for efavirenz, tramadol, tacrolimus, and isoniazid. Pharmacogenomic decision support in African settings must incorporate population-specific allele definitions and input-format-aware strategies.

Routine childhood growth surveillance is a cornerstone of pediatric care. Growth pattern abnormalities are often early manifestations of chronic disease. Yet subtle abnormalities are frequently underrecognized, leading to diagnostic delays and avoidable morbidity. We introduce SPROUT (System for Pediatric Recognition Of Undiagnosed Trajectories), a generalized, multi-agent large language model (LLM) reasoning system designed to identify a broad spectrum of pediatric growth-related conditions from longitudinal electronic health records (EHRs) earlier than standard clinical practice. Using a large pediatric primary care EHR dataset, we developed and validated SPROUT as a two-stage system. First, a highly specific LLM screener flags concerning longitudinal growth patterns. Second, an Orchestrator module coordinates a multidisciplinary panel of LLM agents to generate a ranked differential diagnosis. To correct systemic reasoning errors, a Trainer module injects meta-knowledge into the panel via a dedicated "Learner" agent. Diagnostic capability was evaluated using a walk-forward, visit-by-visit simulation leading up to the diagnosis date. The SPROUT screener model achieved 98% (83/85) specificity and 28% (9/32) sensitivity on a gold-standard dataset of pediatric primary care patients when evaluated one year before the index date, and 100% specificity and 47% sensitivity when evaluated using longitudinal data up to the day of diagnosis. When applied to 300 control patients (i.e., healthy or undiagnosed), the screener flagged 15. Subsequent expert panel review confirmed high suspicion for undiagnosed pathology in 33% (5/15) of these cases. In chronological walk-forward validation on disease cases, the diagnostic engine identified conditions well before standard-of-care documentation. One year prior to clinical diagnosis, the system achieved sensitivities of 81% for type 1 diabetes mellitus, 56% for pituitary disorders, and 44% for celiac disease. The SPROUT multi-agent system demonstrates the ability to detect a significant portion of latent growth-related pediatric conditions months to years before current clinical standards while minimizing false positives. These results support its potential as a decision support tool for reducing diagnostic delays in pediatric care.

Background and Objectives: Deep sequencing of brain tissue in the research setting has established that mosaic variants are a major cause of malformations of cortical development (MCDs) and epilepsy. However, genetic testing in the clinical setting primarily detects germline variants using clinically accessible samples. We aimed to determine the diagnostic yield and clinical utility of deep sequencing in the clinical setting to identify pathogenic mosaic variants for this population. Methods: We performed a retrospective cohort analysis of individuals at Boston Children's Hospital with MCDs with or without epilepsy who received clinical deep sequencing between September 2017 and February 2026. Demographic, clinical, and genetic testing data were abstracted from the medical record. For individuals without systemic features, we classified brain tissue as an affected tissue sample. For individuals with systemic features, we classified brain or relevant non-brain tissue as affected. The primary outcome was the diagnostic yield of clinical deep sequencing performed using affected vs unaffected tissue samples. The secondary outcome was the clinical utility of genetic diagnoses. Results: Our cohort included 37 individuals (19/37 (51%) female, 18/37 (49%) male) with MCDs, of whom 35/37 (95%) had epilepsy (25 with brain tissue samples available from epilepsy surgery) and 8/37 (22%) had systemic features. Most (35/37 (95%)) had dysplasia phenotypes on MRI and 12/27 (44%) with pathology available had Focal Cortical Dysplasia Type I or II. The diagnostic yield was 53% (17/32; 16 mosaic and 1 germline variant) when clinical deep sequencing was performed using an affected tissue sample vs 0% (0/6) using an unaffected tissue sample (p=0.016). Of the diagnosed cases, 13/17 (76%) had testing performed on brain tissue (1 with systemic features) and 4/17 (24%) on non-brain tissue (3 buccal and 1 duodenal tissue, all with systemic features). All but one diagnosis involved the mTOR pathway. All diagnoses had clinical utility. Discussion: Clinical deep sequencing, when performed using an affected tissue sample, has high diagnostic yield and clinical utility for individuals with MCDs, especially dysplasia phenotypes, and epilepsy. Our findings support implementation of clinical deep sequencing for this population, especially as the genetic diagnoses have implications for emerging precision therapies.

Obesity is a significant public health concern. Early-onset obesity in the context of rare disease can reflect genetically-mediated pathology or elevated susceptibility through indirect mechanisms. Mapping the diverse characteristics and needs of young people with obesity in the rare disease population is a first step toward mechanistic and translational research. We carried out a retrospective comparative analysis of demographic, genotypic, phenotypic and health service utilisation data for young people with obesity (cases: n=500) and without obesity (controls: n=11,444) from the UK 100,000 Genomes Project rare disease cohort. Cases and controls were recruited prior to genomic diagnosis, across clinical disorder categories. We observed significant association between socioeconomic deprivation and obesity risk. Young people with obesity had significantly higher utilisations of acute care and mental health services, indicating an overall higher health burden. A curated panel of 519 candidate obesity-associated genes demonstrated aggregate association with obesity, although no single gene reached significance. Phenotypic comparison between cases and controls highlighted increased multi-organ and neurological system involvement, highlighting the overlap between neurodevelopmental and obesity risks. Within the case group, we conducted cluster analysis to identify early-onset obesity groups with different phenotypic profiles, potentially arising from different causal pathways - this identified six obesity subgroups of interest, with differing involvement of neurodevelopmental and other systems. Our study confirms that obesity co-occurs with a wide range of factors within the rare disease population, and is associated with significant physical and mental health needs, requiring holistic lifelong care.

Pathogenic variants in RUBCN, encoding the Run domain Beclin-1 interacting and cysteine-rich domain-containing protein (Rubicon) have been implicated in autosomal recessive spinocerebellar ataxia 15 (SCAR15). However, the molecular mechanisms underlying disease pathogenesis remain poorly understood. Here, we report 18 individuals from 15 unrelated families harbouring biallelic RUBCN variants, who present with an aggressive neurodevelopmental disorder variably characterized by seizures, developmental delay, intellectual disability and movement abnormalities that cause regression, progressive brain atrophy and neurodegenerative features. Through functional characterization, we demonstrate that a subset of disease-associated putative truncating variants disrupt autophagy regulation. In Caenorhabditis elegans models, loss-of-function RUBCN variants result in an increased autophagic flux and impaired neuronal function, recapitulating key features in humans. Correspondingly, cellular assays reveal that nonsense and frameshift RUBCN variants lead to defective autophagy inhibition, underscoring a crucial role for RUBCN as a key negative autophagy regulator. Molecular dynamics simulations rank the eleven missense variants by structural effect, with p.Arg813Trp alone altering the target protein at both the local and the regional level and lying within the RAB7A-binding module that the truncating alleles remove altogether. Our findings establish and expand the RUBCN-related disorders as a clinically and molecularly distinct subset of autophagy-related diseases. By delineating both the genetic landscape and cellular consequences of Rubicon dysfunction, this study enhances our understanding of autophagy-related neurodevelopmental disorders and provides a foundation for future therapeutic investigations.

Genotype imputation remains essential for large-scale human genetics studies, but its performance is limited by the size and ancestral diversity of available reference panels, reducing accuracy for rare variants and underrepresented populations. Here, we present a cloud-based imputation service built on a multi-ancestry reference panel derived from 515,579 jointly phased genomes from the All of Us (N=414,830) and National Human Genome Research Institute's Analysis, Visualization, and Informatics Lab-space (AnVIL, N=100,749) datasets. The All of Us + AnVIL reference panel is highly diverse and includes 261,163 participants most genetically similar to non-European reference populations, spanning 665,398,839 high-quality autosomal sites, representing a nearly 50% increase over TOPMed, the previous largest imputation service. Across multiple ancestry groups, the panel enables accurate imputation (empirical R2 0.8) for variants with allele frequencies as low as 0.2%, extending reliable imputation into the rare-variant frequency spectrum, including allele frequencies down to 0.002% and 0.006% for samples with European ancestry and African ancestry in the United States, respectively. Compared with TOPMed, the panel improves imputation accuracy across all ancestry groups except Africans, and recovers additional trait-associated variants not represented in existing reference panels. To facilitate broad community access while preserving participant privacy, we deploy the panel through a secure cloud-based imputation platform using privacy-preserving recombined haplotypes. This resource establishes a new foundation for genome-wide association studies (GWAS) and fine-mapping, especially in previously underrepresented populations.

Predicting disease risk from prior diagnoses is fundamental to clinical decision-making, particularly during health emergencies such as the COVID-19 pandemic, when individuals with long-term conditions may be disproportionately vulnerable to adverse outcomes. Despite intense interest in developing models to predict disease risk from prior diagnoses (1-3), most prediction models do not estimate effects of each prior diagnosis on disease risk conditional on other diagnoses, limiting interpretability and clinical utility. We developed the Comorbidity Risk Score (CRS), trained on 13 million individuals (age 40-69) from linked electronic health record (EHR) datasets of the entire population of England, to predict COVID-19 hospitalisation and 87 other disease outcomes. CRS was trained at close to saturated sample size and precisely estimated the effects of 212 prior diagnoses on the 88 disease outcomes, conditional on all other prior diagnoses. Correlations of CRS effect sizes across outcomes (e.g. 0.76 for myocardial infarction vs. hyperlipidaemia) matched the corresponding genetic correlations (e.g. 0.79 for myocardial infarction vs. hyperlipidaemia), confirming that comorbidity architectures capture disease aetiology. On average, CRS identified 5% of the population with 3.4-fold higher disease risk, including myocardial infarction (4.4-fold), lung cancer (6.5-fold), and COVID-19 hospitalisation (6.3-fold). Using prior diagnoses alone, CRS outperformed state-of-the-art clinical COVID-19 models (4). Furthermore, CRS (N=13 million) substantially outperformed state-of-the-art AI (1) (N=0.5 million) and linear (3) (N=0.5 million) models in predicting disease risk, suggesting that training sample size outweighs model complexity. CRS attained near-perfect transferability across self-reported ethnicities (e.g., Black vs. White: AUROC ratio = 97.3%). Finally, CRS distinguished independently predictive comorbidities from indirect associations, e.g., lipid metabolism disorder was a strong predictor of myocardial infarction risk but not ischaemic stroke, after conditioning on other prior diagnoses. In conclusion, CRS provides a comprehensive resource for understanding the impact of comorbidities on COVID-19 and other future diseases, revealing disease aetiology while enabling powerful prediction of disease risk.

Standard polygenic risk scores (PRSs) are constructed based on additive genome-wide association study (GWAS) summary statistics. Nonlinear machine learning methods have been increasingly applied to construct PRSs directly from individual-level data, with the aim of improving predictive performance over standard PRSs through their ability to model non-additive genetic effects. However, their superiority across studies has been inconsistent, and the conditions under which they provide meaningful improvements remain unclear. We combined theoretical analysis, simulations and a real-world application to investigate when two widely used nonlinear machine learning methods, random forest and XGBoost, outperform standard PRSs. Theoretical analysis showed that standard PRSs can implicitly capture part of the genetic variance attributable to nonadditive genetic effects through their contributions to marginal SNP effects, thereby losing less information than commonly assumed. Although nonlinear models have a higher theoretical potential, their greater flexibility incurs a bias-variance trade-off that can limit predictive gains at finite sample sizes. Simulations showed that XGBoost outperformed the standard PRS only when the genetic architecture involves a sufficiently large proportion of interaction genetic variance concentrated across relatively few interaction effects and large training samples were available. Random forest consistently underperformed the standard PRS. In an application to ischemic heart disease prediction using UK Biobank data, XGBoost showed no meaningful improvement in predictive performance over the standard PRS, whereas random forest again performed worse. Together, these findings suggest that nonlinear machine learning do not uniformly outperform standard PRSs; rather, their relative performance depends jointly on genetic architecture and training sample size. Our study helps to reconcile the inconsistent results reported across previous studies and provides a framework for identifying settings in which more complex PRS models are likely to be beneficial.

Background. Routine service databases are attractive sources of training labels for clinical prediction models, but the processes that write those labels are rarely audited before the labels are used. In a deployed community cognitive-screening programme, we audited the routine cognitive-status label, built a matrix of twenty-four model arms over the same patients under a specialist reference standard, and measured what each supervision choice bought or cost. Methods. The study cohort is the 672 individuals whose cognitive status was recorded by a titled (attending-or-above) physician, that record being the reference standard; after holding out one institution entirely, a development panel of 642 individuals at 38 institutions. The routine cognitive-status label these individuals also carry was first audited at the operator level: for each data-entry account we counted diagnoses entered and the proportion recording any impairment, and tested a competing bulk-timestamp explanation. Twenty-four arms span the supervision choices such a programme faces: an incumbent 21-variable logistic regression; local language models (Qwen2.5-1.5B/3B, Qwen3-4B/8B) zero-shot, with chain-of-thought, fine-tuned on physician labels, on routine labels with and without decontamination, or on a proxy scale-band task; preference-optimised (DPO) and reinforcement-trained (GRPO) variants; a proprietary frontier model queried zero-shot; and knowledge distillation of that frontier model into the regression and into the local 4B, using 943 teacher-labelled records from the programme's unlabelled pool. All arms are scored out-of-fold under one five-fold split grouped on registry-resolved institution clusters (no cluster spans a fold); paired contrasts use a 2,000-draw cluster bootstrap. Results. 181 operator accounts (each entering at least 100 diagnoses with zero recorded impairments) account for 45,315 rows - 40.5% of the outcome column; recorded impairment falls monotonically with account volume (15.7% for 1-9 rows to 0.7% for 500-999); a bulk-timestamp explanation was tested and refuted, identifying the write-time column as a migration artefact. Under the specialist standard, no locally fine-tuned arm beat the incumbent regression (AUROC 0.926): physician-label SFT reached 0.924 (4B), DPO 0.881, and GRPO 0.789; the pre-registered two-stage proxy-then-RL recipe was worse than its single-stage contaminated baseline (-0.030, 95% CI -0.077 to -0.004). Chain-of-thought reduced discrimination at every size (-0.072, -0.080, -0.041 at 1.5B/3B/4B; -0.012, n.s., at 8B). The frontier model scored 0.932 (vs. regression +0.007, n.s.). The distilled 4B reached 0.940 - above the incumbent (+0.014, 0.004 to 0.031) and above its own teacher (+0.008, 0.001 to 0.017) - with near-teacher calibration; it reached the teacher's level by 50 teacher labels and changed little beyond 200. Conclusions. The audit and the arm matrix support one deployment recipe: audit the routine label at the operator level before training on it; do not expect fine-tuning, preference optimisation, or reinforcement learning on a few hundred specialist cases to beat a well-calibrated regression; and if a frontier model is available but undeployable, spend a bounded number of queries on it as a labelling instrument and distil. A companion paper uses these frozen predictions to quantify how evaluation design choices compare with model choice.

Active Feature Acquisition (AFA) adaptively selects which diagnostic test to order next and offers a route to reduce unnecessary laboratory testing in acute care. Existing clinical AFA evaluations, however, assume every feature can be retrieved on demand and split data at the visit level, both of which inflate apparent performance. We re-evaluate cost-aware AFA under constraints designed to reflect deployment. From MIMIC-IV we constructed a cohort of 64,766 acute admissions (39,884 patients; 21 conditions; 55 features in 30 test panels) with a patient-level split, a 12-hour decision cutoff, and a per-patient availability mask from what was actually measured, and priced panels using the 2026 Medicare fee schedule under panel-level billing. We evaluated EIG-Cost, which scores each panel by Monte-Carlo Expected Information Gain penalised by its dollar cost, against eight published methods across budgets \30--$60 over five patient-level resamples. At a $30 budget, EIG-Cost achieved the highest macro-F1 (0.188, 95% CI [0.185, 0.191]) at the lowest cost ($17.28), exceeding the strongest baseline in all five resamples (p<0.001; Cohen's d=4.0), and led at every budget. Three of the eight methods collapsed to a vitals-only baseline (macro-F1 approx 0.040), acquiring nothing even at higher budgets, a genuine failure to adapt to availability rather than a budget limitation. Despite modest absolute accuracy, EIG-Cost's probabilities were well-calibrated (expected calibration error $0.048$). Under realistic availability constraints, clinical AFA is substantially harder than full-availability benchmarks imply, several published methods fail outright, and cost-aware information-gain scoring is a robust choice in this harder setting.

Over the course of a lifetime, somatic mutations accrue in normal human cells, causing variation in cell phenotype and engendering somatic evolution with outcomes ranging from the adaptive immune system to cancer. To inform understanding of somatic evolution in the human body we report the mutation rates and mutational signatures of 53 normal cell types. Most show evidence of linear mutation accumulation over time with single base substitution mutation rates ranging from ~3.5/year/diploid genome in spermatogonia and sperm, to ~20/year in postmitotic neurons, ~50/year in mitotically active colorectal epithelial cells, ~60/year in kidney proximal tubule cells and hepatocytes, 100s/year in sun-exposed skin epidermal cells and 10-50/year in the remainder. Certain cell types, including skin epidermis, cardiac myocytes, bladder urothelium, kidney proximal tubule cells, and hepatocytes, show substantial variability in mutation burdens around the linear age trend, indicating the influence of additional factors which differ between individuals and modulate mutation accumulation, including exogenous mutagen exposures. At least 18 single-base substitution and nine small insertion and deletion mutational signatures are present, some in all cell types, some in a subset and others in a single cell type. Known exogenous mutagen exposures and endogenous mutational processes account for some mutational signatures, but the origins and mechanisms underlying many are uncertain. This comprehensive survey of mutagenesis provides a foundation for understanding somatic evolution of human cell populations in health and disease.

Agentic workflows can coordinate modelling, but balancing predictive performance, measurement burden and reproducibility is unclear. We developed DXA Agent, an agentic workflow for dual-energy X-ray absorptiometry (DXA) outcomes integrating planning, feature-model refinement, tools, provenance and hypothesis-generating interpretation. Models were independently developed and tested in UK Biobank (5,318 participants) and the National Health and Nutrition Examination Survey (NHANES; 3,777 participants), using cost-efficient and no-limit strategies. Across 20 UK Biobank and three NHANES bone mineral density sites, cost-efficient models achieved lower RMSE and higher R2 than the best conventional comparator, with median relative RMSE reductions of 10.9% and 9.9%, respectively. Classification was task dependent: UK Biobank osteoporosis averaged AUROC 0.839 and PR-AUC 0.182, whereas NHANES performance was comparable with conventional models. Higher-burden features did not consistently improve prediction. These retrospective, cohort-internal findings position DXA Agent as an inspectable, measurement-burden-aware research workflow requiring independent prospective validation.

Unmeasured confounding threatens causal inference and replicability in observational multi-omic studies across variable environments. Genetic instrumental variables (Mendelian randomization) and negative-control calibration each address complementary sources of unmeasured confounding, yet no existing framework unifies them for omics-scale mediation analysis. We introduce ICONIC, an R package that embeds genetic instruments and negative controls within a proximal causal inference framework for total-effect and mediation analysis. ICONIC implements eight estimators spanning five confounding-control strategies, supports continuous, binary, and time-to-event outcomes, and provides extensive diagnostics including sensitivity analyses that map estimator performance across plausible assumptions. Ground-truth benchmarks are calibrated to real-omics covariance structures via a hybrid generative model (GAN + feature-level Gaussian copula) rather than parametric simulation, and a companion planning tool predicts performance gains from collecting additional omic data. We demonstrate ICONIC in two case studies: identifying placental transcriptomic mediators of gestational diabetes on birth weight (n = 164), and tumor-expression mediators of smoking intensity on lung cancer survival (n = 494). Notably, ICONIC's diagnostics recommended different estimation strategies across the two scenarios, reflecting differences in the likely influence of unmeasured confounding. ICONIC is freely available at https://github.com/sbresnahan/iconic/.

Autosomal dominant gain-of-function mutations in platelet-derived growth factor receptor beta (PDGFRb) cause overgrowth of the skeleton and other connective tissue in Kosaki overgrowth syndrome. However, the target cell type and signaling pathways underlying PDGFRb-driven overgrowth are unknown. Normal postnatal growth is controlled by pituitary-secreted growth hormone (GH), which activates the STAT5 transcriptional factor to upregulate insulin-like growth factor 1 (IGF1). To investigate the role of the GH-STAT5-IGF1 pathway in PDGFRb-related overgrowth, we generated mice with a PDGFRb gain-of-function mutation in skeletal and fibroblast lineages, which resulted in STAT5 activation and gigantism. Conditional deletion of Stat5ab in connective tissue lineages rescued skeletal overgrowth and keloid-like fibrosis in the skin. Conditional deletion of GH receptor (Ghr) did not rescue overgrowth, indicating the physiological activator of STAT5 is not required for overgrowth. However, deletion of Igf1, the STAT5 target gene, and its receptor, Igf1r, in connective tissue, rescued the overgrowth phenotype. These findings demonstrate a GHR-independent STAT5-IGF1 signaling pathway in mutant connective tissue cells, which mediates PDGFRb-driven overgrowth in mice and potentially in humans with similar PDGFRB mutations.

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia and a major cause of cardioembolic stroke. Although polygenic risk scores (PRS) are well characterized to quantify inherited susceptibility for AF, they provide limited insight into the pathways and tissues underlying genetic risk, which are critical to uncover for individual risk prediction. In this study, we develop a pathway-level multi-omics representation learning framework that converts individual genetic profiles into interpretable biological features by integrating GWAS-derived pathway burden scores with tissue-specific transcriptomic pathway signals. We constructed machine learning models to assess population-level AF risk prediction performance across genomic and transcriptomic tissue contexts; the pathway-based global attention models substantially improved risk prediction performance over PRS and other baselines (AUROC improved from 0.601 to 0.738). Transformer and graph neural network frameworks then assessed individual-level pathway interpretability, revealing heterogeneous contributions from electrical signaling, cardiac development, and DNA repair pathways to AF risk. This added interpretability highlights the potential of this pathway approach to enable more mechanistically informed risk stratification than static PRS by capturing underlying heterogeneity. To independently assess whether prioritized pathways reflected cardiac regulatory biology, we compared pathway rankings with transcriptional effects predicted by the AlphaGenome foundation model. Variants in highly ranked pathways showed significantly greater predicted effects on expression in atrial and ventricular tissues (FDR = 0.032) relative to controls, providing orthogonal evidence that the model identifies biologically relevant mechanisms. Overall, this work reframes polygenic risk from a single measure of susceptibility to tissue-informed pathway mechanisms, providing a framework for interpretable genomic stratification in complex diseases.

Background: Traditional real-world studies comparing SGLT2 and DPP4 inhibitors on renal outcomes rely on propensity score matching, which causes high-dimensional data loss. We used causal machine learning (Causal ML) to unmask heterogeneous treatment effects in diabetic kidney disease (DKD). Methods: Using data from 4,588 patients within the Japanese J-CKD-DB-Ex registry, we implemented a doubly robust (DR) learning framework (Linear DR-learner with XGBoost) to compare SGLT2 and DPP4 inhibitors. Outcomes included the chronic eGFR slope and a composite renal endpoint ([&ge;] 50% eGFR decline or end-stage kidney disease). Heterogeneity was explored via causal SHAP and decision trees. Results: At the population level, SGLT2 inhibitors modestly slowed chronic eGFR decline (average treatment effect [ATE] = 0.14 [95% CI: -0.86, 1.15] mL/min/1.73m^2/year) and reduced composite endpoint risk by 9% (ATE: -0.09 [-0.11, -0.08]) versus DPP4 inhibitors. However, individual-level counterfactual analysis suggested that for the chronic eGFR slope, non-glinide users with stable pre-treatment trajectories who were also taking ACE inhibitors had a greater benefit from SGLT2 inhibitors (ATE: 2.95 [-0.68, 6.58]). Conversely, glinide users with steep pre-treatment decline had a greater benefit from DPP4 inhibitors (ATE: -8.98 [-16.11, -1.85]). For composite renal events, SGLT2 inhibitors had a 28% absolute risk reduction within the algorithmically identified high-risk subgroup (eGFR [&le;] 28.1 mL/min/1.73 m^2 and positive proteinuria; ATE: -0.28 [-0.33, -0.23]). Even non-proteinuric decliners demonstrated a 8% risk reduction with SGLT2 inhibitors (ATE: -0.08 [-0.10, -0.06]). Conclusion: Causal ML advances precision medicine in DKD, shifting from uniform prescribing to individualized, data-driven therapy targeting distinct intrarenal pathways.

Background. The imprecision of current metrics to capture the complex genetic admixture and racial identity among Hispanic/Latino individuals in the United States [US] is a concern. We examined the relationship of self-reported race and genetic ancestry with hypertension [HTN] among Hispanics/Latinos. Methods. Cross-sectional study of the Hispanic Community Health Study/Study of Latinos (HCHS/SOL), including 10,586 Hispanic/Latino unrelated adults. Genetic ancestry: West African [AA], Amerindian [AI], and European [EA]. Self-reported race: White, Black, Native American, or Multiple/Missing (More than one race or Unknown/Not reported/Refused). HTN: systolic (SBP) [&ge;]130 mmHg, diastolic blood pressure (DBP) [&ge;]80 mmHg, and/or use of HTN medications. Age- and sex adjusted models were used. Results. Self-reported race was White (38{middle dot}6%), Black (3{middle dot}6%), Native American (4{middle dot}1%), and Multiple/Missing (53{middle dot}7%), with Unknown/Not reported/Refused representing 32{middle dot}7%. Black and White Hispanics/Latinos had the greatest AA (55{middle dot}7%) and EA (69{middle dot}3%) ancestries, respectively. Each 10% AA increase was associated with OR 1{middle dot}15, SBP beta +0{middle dot}9 mmHg, and DBP beta +0{middle dot}7 mmHg. Conversely, each 10% AI increase was associated with OR 0{middle dot}83, SBP beta -0{middle dot}4 mmHg, and DBP beta -0{middle dot}6 mmHg. HTN prevalence was highest among those with Black race or in the highest AA quantile (45{middle dot}6% and 48{middle dot}0%, respectively), and lowest among those with Native American race or in the highest AI quantile (37{middle dot}6% and 26{middle dot}7%, respectively). Conclusion. One-third of Hispanics/Latinos did not self-report race. Black or White self-reporting race did somewhat relate to AA or EA ancestry, respectively. HTN profiles were related to self-reported race and genetic ancestry in this admixed population.

Recent large-scale studies have enabled new knowledge about genetic underpinnings of morphological and electrophysiological alterations of the retina. Variation in retinal traits, often of neurodevelopmental origin, have been linked to major psychiatric disorders (MPDs). Here, we investigate the genetic overlap between MPDs and key retinal traits to identify underlying molecular mechanisms. We obtained genome-wide associations studies data for bipolar disorder (BD), major depression (MD), schizophrenia (SCZ), and the retinal traits retinal nerve fibre layer thickness (RNFL), ganglion cell inner plexiform layer thickness (GCIPL), and vertical cup-disc ratio (VCDR). We estimated the number of trait-influencing variants shared between traits with MiXeR and identified shared genetic loci with condFDR. Subsequently, we examined the biological pathways of the genes mapped to shared loci. This revealed that GCIPL shared the most genetic variants with MPDs (~60%), followed by RNFL (~40%), and VCDR (~20%). The genetic variants shared between retinal traits and MPDs showed disorder-specific patterns with more pronounced overlaps of SCZ and BD with RNFL, and MD negatively correlated with GCIPL. Gene-pathway analysis highlighted the importance of GABAergic neurotransmission and a two-stage neurodevelopmental process in SCZ, whereas the role of mitochondria and a weaker developmental component were observed in BD. The results also implicated synaptic functioning and gene-expression processes in MD. Furthermore, polygenic analysis suggested that the genetic architecture of retinal traits can distinguish between MPDs. Our findings indicate shared genetic underpinnings between retinal traits and SCZ, BD, and MD, implicating altered neurodevelopment and neurotransmission underlying the retinal link to major psychiatric disorders.

Artificial intelligence applied to routine electrocardiograms (ECGs) has largely focused on detecting existing disease or predicting individual cardiovascular outcomes. Whether ECGs can support prediction of multiple future diseases across organ systems remains unclear. We developed ECG-RISK, a multitask survival model for 67 incident three-character ICD-10 endpoints using ECG waveforms, demographic characteristics and routinely collected laboratory data from 86,673 MIMIC-IV patients. Discrimination was highest for heart, brain, kidney and lung endpoints, with organ-level C-indices ranging from 0.796 to 0.825, whereas liver and pancreatic endpoints showed lower discrimination. The ECG-only model achieved strong discrimination across most endpoints, whereas the incremental improvement gained by incorporating ECG and laboratory inputs beyond demographic information varied substantially across endpoints. Across the nine exploratory aggregated outcomes, Kaplan Meier curves showed clear separation among model-score tertiles. Discrimination was highest for dementia (C-index, 0.891) and heart failure (C-index, 0.857). These findings support the feasibility of ECG-based longitudinal risk prediction across multiple diseases. External validation and competing-risk analyses are required to assess generalisability and clinical utility.