Human Genetics and Genomics Advances

Polygenic scores (PGSs) have progressed toward clinical utility in recent years, demonstrating odds ratios that are comparable with both lifestyle-related and monogenic risk factors for several complex diseases. However, PGS implementation relies on developing analytically and clinically valid assays. Methods for developing polygenic models (PGMs) that incorporate tens of thousands to millions of variants require either genome-wide chip genotyping combined with imputation or whole genome sequencing, limiting their portability. Additionally, variants may be in regions where the local genomic sequence makes it difficult to assay, and multi-allelic or strand ambiguous variants can make bioinformatics more complicated. Because of these challenges, attempts to develop a reliable assay for published PGMs can produce multiple variations from the same underlying model, hindering routine clinical deployment and incorporation into multi-factorial risk prediction models. In this analysis, we present an adjustment to the previously published select and shrink with summary statistics (S4) PGM method designed to optimize performance for realistic clinical use by restricting to a pre-specified number of available SNPs that are well-captured by genotyping arrays. We compare the performance of this oligogenic S4 method (S4-Oligo) to currently published PGSs as well as the full S4 model for prediction of breast, colorectal, ovarian, and prostate cancers in the All of Us Research Program, Biobank Japan, and UK Biobank. We find that S4-Oligo matches or exceeds the accuracy of published PGMs at comparable model sizes while retaining between 88% and 99% of the performance of the original S4 model when limited to 1,000 variants.

Sequence models predict the molecular effects of genetic variation, yet their contribution to interpreting Alzheimer's disease (AD) association signals remains uncertain. Here we analysed AlphaGenome Atlas predictions for 111,446 common variants across 75 regions from a European-ancestry genome-wide association study (GWAS) of AD and related dementias. Predicted expression effects modestly tracked association within loci (mean partial Spearman {rho} = 0.018, P = 0.003) and were higher in variants with the highest posterior inclusion probability than in correlated variants outside credible sets (mean difference, 0.044 percentile units; P = 0.005). This relationship remained after adjustment for genic position ({rho} = 0.011, P = 0.026). Predicted splicing effects also tracked association, but this relationship was attenuated after the adjustment. Expression and splicing scores were associated with the corresponding Genotype-Tissue Expression (GTEx) quantitative trait locus (QTL) status, although these analyses did not establish colocalization or agreement in allelic direction. Across 50 loci with multiple fine-mapping candidates, predicted cellular labels were no more concentrated than in position-matched comparisons, and expression direction profiles showed no detectable excess concordance with lead variants. These findings support expression predictions as complementary evidence for interpreting AD risk variants, while showing that the assignment of cellular context must account for fine-mapping uncertainty.

Polygenic scores (PGS) are increasingly common in human genetics research, yet their performance and portability vary substantially across traits and populations. We systematically evaluate prediction accuracy and cross-ancestry portability for 3,994 published PGSs across the phenome in the ancestrally diverse CCPM biobank. We identify widespread trait-specific and cross-trait associations, demonstrating both the utility and breadth of existing polygenic predictors. However, predictive effects frequently differ across ancestry groups, highlighting persistent challenges for portability that threaten equity in downstream applications. We further find that accuracy and portability depend strongly on whether the PGS is continuous or thresholded. Finally, through hierarchical mixed modeling, we identify score-development, genome features, and cohort characteristics associated with improved PGS accuracy and portability. These results highlight key factors supporting future development of robust and portable scores for research and clinical translation.

Polygenic scores (PGS) predict an individual's genetic value for a complex trait of interest by aggregating effect size estimates from genome-wide association studies (GWAS). While powerful, this approach assumes that genetic variants affect the phenotype in an additive way, which might be problematic when dominance is part of the genetic architecture. To address this limitation, we developed PGSa,d, a simple polygenic scoring framework, rooted in quantitative genetics theory, that incorporates both additive (a) and dominance effects (d) directly. We evaluated PGSa,d against the standard GWAS-based PGS (PGS{beta}) in simulations. The results showed that PGSa,d consistently achieved lower root mean squared error (RMSE) and progressively higher R2 as dominance contributed more to the genetic architecture. We then applied the method to 16 blood cell traits, height, and body mass index (BMI) from UK Biobank. PGSa,d showed comparable R2 to PGS{beta} for all traits (including height and BMI whose architecture is predominantly additive) and significantly lower RMSE for four traits after strict Bonferroni correction. Across traits, RMSE improvement from PGSa,d was positively associated with the magnitude of the estimated dominance effects, whereby traits with larger dominance effects displayed lower RMSE from PGSa,d. As a byproduct of our analysis, we also identified 141 variants in 39 genomic regions with significant dominance effects after Bonferroni correction across traits, recapitulating known associations and identifying novel ones. Together, these results show that explicitly modeling additive and dominance effects is a robust strategy that can improve individual-level predictions, especially when dominance contributes to complex trait architecture.

Background: Triple-negative breast cancer (TNBC) disproportionately affects women of African ancestry, yet the biological mechanisms underlying this disparity remain poorly understood. While recent genome-wide association studies (GWAS) have identified numerous TNBC risk variants, their functional consequences have not been systematically evaluated in African-ancestry populations. We hypothesized that ancestry-specific regulatory variants drive distinct transcriptomic and proteomic landscapes that contribute to TNBC aggressiveness. Methods: We extracted 2,849 unique TNBC-associated variants (p-value [≤] 0.0001) from the GWAS Catalog. Of these, 206 were excluded due to absence in our whole-genome sequencing (WGS) dataset or low minor allele frequency ([≤] 0.01). We prioritized the remaining 2,643 bi-allelic variants for analysis in the GENE-FORECAST cohort. We integrated WGS with whole-blood transcriptomic (n=242) and proteomic data (n=362) to perform cis-eQTL and cis-pQTL analyses in discovery and validation sets. Results: We identified 38 cis-eQTL associations and 5 cis-pQTL associations that were statistically replicated across independent discovery and validation datasets within our cohort. Both omics layers converged on a single gene, Hemoglobin Zeta (HBZ). We identified five functional variants (four SNPs and one INDEL) in strong linkage disequilibrium that consistently upregulate HBZ mRNA and the corresponding circulating embryonic zeta-globin protein. These variants were found almost exclusively in African and African-admixed populations across GENE-FORECAST, 1000 Genomes, BioVU, and All of Us datasets. Four of the five variants were also confirmed as significant HBZ eQTLs in GTEx whole blood. Conclusion: Our integrative analysis reveals variants of the embryonic HBZ gene as a novel, ancestry-specific TNBC-associated risk factor that is detectable at both transcriptomic and proteomic levels. This suggests that aberrant reactivation of this embryonic globin may provide a selective advantage in the hypoxic TNBC microenvironment. These results highlight a unique mechanism of tumorigenesis in women of African ancestry, offering a potential distinct biomarker and therapeutic target for this high-risk population.

Multiplexed assays of variant effect (MAVEs) and computational methods have advanced variant classification but have struggled to distinguish the molecular mechanism of variants with sufficient accuracy, limiting utility for multi-mechanism genes. Here, we coupled single-cell RNA-seq profiling with supervised machine learning to predict the mechanism of variants in CASR, a gene in which gain of function (GOF) variants cause hypocalcemia and loss of function (LOF) variants cause hypercalcemia. We engineered endogenous CASR-depleted HEK293 cells to express a single exogenous variant copy, measured their transcriptional profiles at single-cell resolution, trained a multi-class machine learning model on 96 expert-annotated variants (25 benign, 35 LOF, 36 GOF), and predicted mechanism for 157 variants of uncertain significance, of which 16 were predicted GOF and 54 LOF. The model showed high performance (macro F1 0.98, overall accuracy 0.98), and its predictions were corroborated by clinical phenotype and serum calcium measurements across more than 300,000 patients in a commercial genetic testing database. These results highlight the effectiveness of using molecular mechanism annotations with scRNA-seq to train models that distinguish molecular mechanisms of pathogenicity. The approach has immediate relevance for clinical variant classification and establishes a foundation for mechanism-guided clinical management in disorders with mechanism-dependent interventions, with potential application to other multi-mechanism genes.

Genetic prediction of clinically defined disease subtypes could advance precision medicine, but polygenic risk scores (PRSs) often show limited ability to distinguish subtypes, either because the subtypes share genetic architecture or because available PRSs are inaccurate. We introduce GenSep, a statistical framework that estimates the oracle case-case AUC attainable if the true additive genetic values of both subtypes were known, and the proportion of this genetic separation recovered by subtype-specific PRSs. Across 18 subtype pairs in UK Biobank, oracle AUCs ranged from 0.605 to 1.000, whereas current PRSs achieved a median observed case-case AUC of only 0.530 (range, 0.510-0.754), corresponding to recovery of a median of 4.1% (range, 0.3-15.5%) of the genetic separation variance available to a perfect predictor. Oracle AUCs were generally consistent across UK Biobank, All of Us, FinnGen and the Million Veteran Program, and across European, African and Admixed American ancestry groups. FinnGen-trained PRSs, despite 2.6-to 3.9-fold larger discovery samples, performed no better than internally trained PRSs, and substantially larger consortium GWAS raised case-case AUC by 0.06 on average (range, 0.003-0.089). GenSep separates subtype pairs limited by intrinsic genetic overlap from those limited by current PRS accuracy, identifying where larger studies and better predictors could improve genetic discrimination.

Structural variants (SVs) are larger genetic rearrangements such as deletions, duplications, and insertions and other forms of structural alterations. Given their large size, they can have an outsized effect on phenotype, however, the contribution of SVs to the genetic architecture of orofacial clefts (OFCs) is not well understood. In this study, we sought to better understand the contribution SVs to OFC risk by characterizing de novo SVs (dnSVs) in 1,549 probands with an OFC. Overall, we called 375 dnSVs in 307 probands (0.242 dnSVs per genome). The dnSV rate per genome was highest among individuals with cleft lip (CL, 0.29), followed by cleft lip and palate (CLP, 0.25) and cleft palate only (CP, 0.20), though differences across subtypes were not statistically significant. There were also no differences between sex or syndromic status, but we did identify a significant difference in the rate of duplications across OFC subtypes (p=0.005). Compared to CP (0.028), duplication rates were significantly higher for both CL (0.121; p = 0.003) and CLP (0.096; p = 0.002), suggesting a potential etiologic mechanism for CL/P. We systematically curated each dnSV to classify its potential for influencing OFC risk by examining the genes and regulatory elements disrupted by the SV and evidence for their role in craniofacial development or OFCs. We identified 22 dnSVs in regions associated with recurrent genomic disorders, including seven deletions and one duplication of the well-established OFC risk locus 22q11.2. Additionally, we identified 12 large (>1Mb) SVs, with eight meeting criteria for a likely pathogenic classification based on total gene involvement, and seven dnSVs disrupting well-known genes associated with autosomal dominant forms of OFC. The most notable was SATB2, for which we found four dnSVs, all in individuals with CP. Next, by combining evidence from recurrent dnSVs and de novo single nucleotide variants with Hi-C contact maps and predicted chromatin states from developing craniofacial tissues, we found new candidate genes for OFC risk including FHAD1, NBEA, NR3C2, and SPHKAP. Overall, dnSVs classified as likely pathogenic accounted for putative molecular diagnoses in approximately 1.67% of the cohort, with an additional 0.96% having variants of unknown significance with plausible risk for OFCs. Our results support dnSVs as an important component of OFC etiology and suggest their inclusion in analyses may improve diagnostic yield for affected individuals.

ReNU syndrome is a severe neurodevelopmental disorder caused by de novo variants in the spliceosomal small nuclear RNA (snRNA) gene RNU4-2. Pathogenic variants cluster in two distinct structural regions of the U4 snRNA produced by RNU4-2, the T-loop and Stem III. Aberrant 5' splice site selection is a molecular hallmark of ReNU syndrome that correlates with phenotypic severity and could be leveraged as a quantitative biomarker to facilitate preclinical and clinical therapeutic development. We analysed RNA-sequencing data from two independent cohorts totalling 30 individuals with ReNU syndrome and 54 controls. We identified 483 alternative splicing events shared across both cohorts, with the highest concordance in 105 alternative 5' splice-site (A5SS) events. Using these 105 A5SS events, we derived a minimal reproducible signature from only six sites. The splicing signature perfectly distinguished 19 n.64_65insT individuals from 54 controls, and produced no false positives when applied to a further 5,984 whole blood controls. In addition, the signature was specific to variants within the U4-2 T-loop, which are associated with greater clinical severity. Further, we find evidence of alternative splicing events that are specific to variants in Stem III. In a patient-derived induced pluripotent stem cell (iPSC) model of the recurrent RNU4-2 n.64_64insT variant, the splicing signature was fully recapitulated across all stages of differentiation into cortical neuron-like cells. Together, these findings define a robust RNA biomarker for T-loop ReNU syndrome for use in therapeutic development, and illustrate a general framework for portable biomarker discovery in spliceopathies.

CACNA1A-related disorders arise from altered function of the P/Q-type calcium channel Cav2.1 and encompass a broad spectrum of neurological phenotypes, including developmental and epileptic encephalopathy, familial hemiplegic migraine type 1 and episodic ataxia type 2. Although loss-of-function and gain-of-function mechanisms have both been implicated, interpretation of rare CACNA1A missense variants remains challenging because clinical presentation is heterogeneous and many variants identified through diagnostic sequencing remain classified as variants of uncertain significance. To generate clinically interpretable functional evidence, we developed a calibrated high-throughput automated patch-clamp assay to assess the CACNA1A variants. Twenty-six curated benign/likely benign and 30 curated pathogenic/likely pathogenic missense variants were expressed in Flp-In HEK293 cells with CACNB4 and CACNA2D1. We quantified current density, {Lambda}V0.5 activation,{Lambda}V0.5 inactivation and recovery from inactivation (T0.5 RFI), and used benign controls to define parameter-specific normal ranges with {+/-}3 Z-score thresholds. This multi-parameter framework classified all benign controls as functionally normal and 29 of 30 pathogenic controls as abnormal, yielding 100% specificity and 96.7% sensitivity. Current density identified 19 pathogenic controls with loss of function, while gating and recovery from inactivation identified additional abnormal variants with preserved or partially preserved current density. The assay generated likelihood ratios of 0.033 for normal function and 25.13 for abnormal function, supporting BS3 and PS3 functional evidence, respectively. Application to 23 CACNA1A variants identified in Australian patients showed abnormal function in all four variants already classified as pathogenic or likely pathogenic. Among 19 variants of uncertain significance, 11 were functionally abnormal and nine were reclassified as likely pathogenic after integration of PS3 evidence with other ACMG/AMP criteria. These data show that CACNA1A missense variants can produce loss-of-function, gain-of-function or mixed effects, and that multi-parameter functional assessment is required for accurate interpretation. This calibrated automated patch-clamp framework provides clinically interpretable evidence for refining CACNA1A variant classification in diagnostic practice.

Protein coding variants span the gamut of structural and functional effects on proteins, and they have similarly heterogeneous effects on traits. Their functional effects can be predicted and categorized, particularly using machine learning models, and integrating these predictors benefits gene discovery. However, it remains unclear how well known structural features and predicted functional scores explain the heterogeneous effects of protein coding variants on human complex traits. We introduce the Kernel Allelic Series Model (KASM), which integrates multiple functional annotations to form a kernel describing how allelic effects within a gene covary. We apply it to exome data from 26 traits and 786,871 individuals from UK Biobank and All of Us, integrating 35 annotations describing protein structure and predicting allelic effects. We estimate that rare missense heritability is at least 1.27{+/-}0.09% on average, 1.83{+/-}0.11% under our most flexible model, exceeding pLoF heritability (0.87{+/-}0.05%) and far exceeding missense burden heritability as previously estimated (0.54{+/-}0.04%). AlphaMissense is the single most enriched missense predictor, but a composite score doubles its enrichment. A series of structural features is enriched for heritability, from amino acid chemical features and structural motifs to residue contacts and protein-protein interaction interfaces. Examining pairs of variants, we find that most missense effects are modestly correlated, but an exception is when residues are close in 3D structure but distant in 1D sequence; such pairs also covary across species. Finally, we analyze individual genes and identify gene-specific enrichments, like a 95-fold enrichment among PIEZO1 hydrogen bond residues, that nominate functional hypotheses.

As the use of population descriptors such as race, ethnicity, and ancestry have become increasingly common in modern genetics research, there have been growing calls to critically examine their use. Most notably, in 2023, the National Academies of Science, Engineering, and Medicine (NASEM) published a report titled Using Population Descriptors in Genetics and Genomics Research: A New Framework for an Evolving Field, which included eight specific and actionable recommendations for researchers to implement the ethical and accurate use of population descriptors in genetic research. Here, we use the 2023 NASEM report as a benchmark to analyze the use of population descriptors in genome-wide association studies (GWAS). We develop a general toolkit for large language model-based bibliometrics, operationalize the report's recommendations into an evaluation framework, and apply this framework to evaluate all 4,007 papers from the GWAS Catalog published between 2007 and 2025 with full text available on PubMedCentral. We find significant improvements in adherence to NASEM report recommendations over time. However, most improvements predate the publication of the NASEM report itself, suggesting the report functioned primarily as a synthesis of existing best practices rather than a catalyst for change. We conclude by highlighting opportunities for growth in the field of human genetics.

Bipolar disorder (BD) is highly heritable, yet the contribution of rare coding variation remains incompletely characterized. We analyzed sequencing data from 64,435 individuals with BD and 168,101 controls spanning multiple ancestries and 22 countries, representing the largest and most global sequencing resource with a 6.7-fold increase in effective sample size over the previous study iteration. We observe enrichment of protein-truncating and damaging missense variants in constrained genes and curated neuropsychiatric gene sets, with no enrichment of synonymous variation. These enrichment signals were consistent across ancestry groups, suggesting that genetic risk factors for BD are consistent worldwide. Gene-level analyses identified 13 exome-wide significant genes and an additional 20 genes at FDR < 0.05. These genes showed convergence with common and rare variant risk across other neuropsychiatric disorders. Expression analyses also showed preferential brain expression and increased developmental expression during early childhood. Modelling 3D protein structures further highlighted clustering of ultra-rare missense variants at a predicted interaction interfaces in two Bonferroni-significant genes, DOP1A and ATP9A; a potential mechanism linking membrane trafficking to BD risk. Together, these results implicate a constellation of rare variants that map onto neuronal biology and demonstrate that diverse global populations converge on shared genetic signals underlying BD risk.

Ancestry-related genomic variation may contribute to phenotypic diversity. However, its cellular consequences remain difficult to evaluate. To address this, we developed a human induced pluripotent stem cell (hiPSC)-based framework for investigating such phenotypes. We used the Japanese population as a model of Jomon-related and continental East Asian ancestry admixture. Jomon ancestry, estimated using the Jomon allele score (JAS), was associated with clinical measures, including renal-related traits and serum triglycerides in 34,152 Tohoku Medical Megabank participants. Peripheral blood mononuclear cell transcriptomics of donors from the upper and lower extremes of the JAS distribution (n = 10) suggested group-level differences in immune-related gene expression. To test whether the serum triglyceride association coincided with hepatic cellular phenotypes, we differentiated hiPSCs from these donors into hepatocyte-like cells. Neither intracellular triglycerides nor lipid-related gene expression differed between groups. These findings highlight the value of a genetically characterized hiPSC-based framework for investigating ancestry-associated phenotypes across cellular contexts.

Background: The inner ear is supplied by the labyrinthine artery, which is functionally close to an end artery with limited collateral reserve. Microvascular abnormalities have been implicated in sudden sensorineural hearing loss (SSNHL) and age-related hearing loss, yet inner-ear microvasculature cannot be assessed in vivo and current hearing genome-wide association studies (GWAS) are underpowered. We evaluated the shared genetic architecture of retinal arteriolar tortuosity (retina_AT) with blood pressure, cerebral small vessel disease (cSVD) and hearing phenotypes, and examined whether retinal microvasculature could serve as a candidate genetic window to inner-ear microvasculature. Methods: We used linkage-disequilibrium score regression (LDSC) to estimate SNP heritability for 19 phenotypes and genome-wide genetic correlations (rg) for 11 prespecified pairs. We ran approximate Bayes-factor colocalization (coloc.abf) between 19 independent retinal loci and systolic blood pressure (SBP), together with 95% credible-set fine-mapping, UCSC gene annotation, colocalization with GTEx aortic and tibial-artery expression QTL (eQTL) data and with white-matter hyperintensities (WMH), hearing-GWAS power calculations, and univariable and multivariable Mendelian randomization (MR) of blood pressure and heart rate on hearing. Results: Only retina_AT with SBP was significant genome-wide (rg = 0.10, P = 4.2x10, FDR q = 0.00042). The strongest retinal locus, 13q34, did not colocalize (PP.H3 = 0.996; peaks about 267 kb apart; allelic heterogeneity), whereas two secondary loci colocalized strongly--CSK at 15q24 (PP.H4 = 0.98; 3/3 credible variants within CSK) and PAWR at 12q21 (PP.H4 = 0.99; 3/5 within PAWR)--supported by tissue-matched, moderate eQTL evidence (PP.H4 = 0.71-0.75). At 13q34, retina_AT with WMH was an adequately powered null (PP.H4 = 0.019; only this region and phenotype were tested). On the hearing side, sharing of the blood-pressure magnitude (rg {approx} 0.10) was a near-powered null (about 1.5-fold larger sample needed for 80% power under that assumption, versus about 15-fold for weak rg {approx} 0.03 sharing), and rg [&ge;] 0.12 was excluded; MR found no robust causal effect. Conclusions: Retinal arteriolar tortuosity shows limited, distributed genetic sharing with blood pressure at the vascular-tone genes CSK and PAWR; no sharing with WMH was observed at 13q34 (only this region was tested). Retinal microvasculature may serve as a candidate genetic window to inner-ear microvasculature, but this analogy awaits direct testing by larger, vessel-oriented hearing GWAS and inner-ear tissue functional studies.

Introduction: Killer-cell immunoglobulin-like receptor (KIR) gene content shapes natural killer-cell responses and varies substantially among populations. We examined whether worldwide KIR diversity reflects regional structure, isolation by distance, serial founder effects, or latitudinal gradients. Data and methods: Carrier frequencies for 11 variable KIR genes were obtained from the Allele Frequency Net Database. The primary analysis included 189 locally sampled, non-diaspora, non-mixed populations from 70 countries, comprising 25,469 individuals; 161 populations had complete 11-gene profiles. Diversity indices, ordination, clustering, permutation-based variance decomposition, Mantel tests, and regression models were used to characterize geographic structure. Serial founder effects were evaluated using four distance measures from East Africa, both globally and within the Old World. Models of latitude incorporated spatial autocorrelation. Prespecified sensitivity analyses evaluated alternative population sets, gene panels, geographic classifications, weighting schemes, and multivariate outlier exclusion. Results: Geographic region explained 53.5% of variation in KIR profiles, substantially exceeding the permutation-null expectation of 6.9%, although multivariate dispersion differed among regions. Twenty-five of 28 pairwise regional comparisons remained significant after false-discovery-rate correction. Isolation by distance was modest (Mantel r=0.3003) and weakened after controlling for region (r=0.1557). KIR diversity did not decline with distance from East Africa globally but decreased strongly within the Old World (expected heterozygosity {beta}=-0.0066 and standardized gene content {beta}=-0.0789). B-content decreased nominally with absolute latitude, but latitude was collinear with region and the association survived neither spatial modelling nor sensitivity analyses. Regional differentiation remained robust across analytical specifications, whereas the Old World decline was carried by gene content once standard errors were clustered by country. Conclusion: Worldwide KIR gene-content variation shows pronounced regional differentiation, modest isolation by distance, and no independent latitudinal gradient. Diversity declines with distance from East Africa across the Old World, supporting serial founder effects, but not once the Americas and Oceania are included.

Genomic foundation models (GFMs) are pretrained on large genomic datasets, but their ability to predict rare disease-associated events after fine-tuning remains unclear. We tested this using TDP-43 cryptic exons, well characterized splicing events linked to neurodegenerative disease that are poorly represented in standard annotations. We compared the state-of-the-art GFM AlphaGenome with OpenSpliceAI, a specialized splice site prediction model. Fine-tuned OpenSpliceAI detected held-out cryptic exons in human and mouse and captured allele-specific splicing effects at the disease-associated UNC13A locus. In contrast, a splicing prediction head trained on AlphaGenome's frozen pretrained backbone failed to predict cryptic exons. Unfreezing and fine-tuning all model weights, however, enabled AlphaGenome to generalize to held-out data. To investigate cryptic splicing in cell types that are difficult to study experimentally, we applied fine-tuned OpenSpliceAI across the genome to study cell type-specific genes. This led to the identification of new cryptic splicing events in CLDN11 and ARHGAP23, which were subsequently validated in FTLD-TDP brain tissue. Notably, the cryptic exon in CLDN11 is predicted to generate a neoepitope that could serve as an oligodendrocyte-specific biomarker of TDP-43 pathology. Our results show that even a state-of-the-art GFM can fail to generalize to rare splicing events, even with substantial pretraining. However, both foundation and specialized models that learn to generalize to rare biological events can have broad applications in biological discovery and medicine.

Tandem repeat expansions cause more than 60 Mendelian disorders but are difficult to detect and size accurately with short-read sequencing because of their length and repetitive structure. Long-read sequencing (LRS) resolves these loci directly, yet analytic tooling for nanopore sequencing data has lagged behind that available for other long-read platforms. Here we present Mahi-mahi, a tandem repeat genotyper designed for nanopore reads that reports allele-level repeat length, motif composition, interruptions, and read-level flags for systematic basecalling artifacts. Across six Genome in a Bottle samples, whole-genome and adaptive sampling (real-time targeted) data returned calls within one repeat unit of each other for 829 of 832 alleles, and Mahi-mahi calls were within one repeat unit of orthogonal HiFi data at 95.6% of compared alleles after excluding three loci at which the two tools define the repeat differently. In 245 samples from the 1000 Genomes Project, Mahi-mahi calls were within one repeat unit of assembly-derived calls for 98.6% of 10,927 alleles at 57 loci, and Mahi-mahi distinguished pathogenic from nonpathogenic RFC1 alleles of overlapping size by motif. Across 14 Coriell reference lines, Mahi-mahi identified the expected expansion in every line and agreed with orthogonal assays to within two repeat units at HTT, ATXN1, DMPK, and PABPN1. Finally, in 103 samples from three laboratories, Mahi-mahi detected an expanded allele in all samples in which prior testing had reported a pathogenic expansion, assigned the same category as the referring laboratory for 98% of 180 genotypes, and identified known ATXN1 interruption structure. Our results demonstrate that Mahi-mahi accurately identifies repeat expansions at disease-associated loci and could serve as a single assay to evaluate repeat expansions.

Sex differences in human health are typically modeled as binary comparisons between males (XY) and females (XX). However, sex-linked molecular traits vary continuously and are shaped by environmental factors. Here, we introduce the Proteomic Sex Index (PSI), a continuous, plasma proteomics-based measure that quantifies the degree to which an individual's proteome aligns with their chromosomal sex, with higher PSI values indicating profiles less typical for their sex. Leveraging data from 51,310 individuals from six demographically heterogeneous cohorts, we show that PSI captures broad, biologically meaningful variation in sex-enriched traits and diseases. Individuals with proteomic profiles shifted away from their chromosomal sex exhibited trait values less characteristic of that sex. For example, males with higher PSI were shorter, had lower grip strength, and showed reduced severity of COVID-19 (OR=0.59, p=3.0x10-7). Analyses of individuals undergoing gender-affirming hormone therapy showed that, although PSI correlated with circulating sex hormones, it also captured additional variation independent of hormone levels. PSI is partially heritable and genetically influenced. Individuals with Klinefelter syndrome (XXY) had markedly elevated PSI (OR=13.7, p=1.1x10-13), and genome-wide association studies of PSI identified sex-biased cis-pQTLs with evidence of causal links to sex-specific disease risk. PSI provides a nuanced, individualized measure of sex-related biology that enables the study of within-sex molecular variation and its links to health and disease.

Background Short-read genome sequencing generates the signal necessary to detect balanced reciprocal translocations at single-nucleotide resolution, but standard structural variant callers represent them as breakend pairs. Structural variant callers produce thousands of breakend records per genome, the majority of which are false positives arising from segmental duplications, repetitive elements, and reference assembly artifacts. This specificity problem has prevented balanced translocation detection from entering routine genome sequencing workflows. We developed a highly specific pipeline to resolve balanced reciprocal translocations and applied it across five clinically distinct cohorts to characterize the prevalence and diagnostic contribution of balanced reciprocal translocations. Methods Short-read genome sequencing data from 16,131 individuals across five cohorts were analyzed: healthy participants from the 1000 Genomes Project (n=3,202), probands with structural birth defects and their family members (CHOP BDB cohort; n=2,875, including 1017 probands), probands enrolled in the GREGoR consortium (GREGoR cohort; n=3,574 probands), probands with undiagnosed disease and their family members enrolled in the Undiagnosed Diseases Network (UDN cohort; n=6,071 including 2,027 probands) and neonates who underwent rapid genome sequencing (CHOP Baby Eagle cohort; n=409). A computational pipeline was developed to characterize breakend events from standard structural variant callers to high-confidence balanced translocation calls by exploiting well-characterized properties of true balanced translocation signatures in short-read data. Translocation breakpoints were resolved to single-nucleotide precision and interrogated for disruption of known disease genes or potential regulatory consequence otherwise. A subset of the identified breakpoints was confirmed by Sanger sequencing. Results Balanced reciprocal translocations were identified in 0.196% of structural birth defect probands (2/1,017), 0.25% of the GREGoR probands (9/3,574), 0.39% of the UDN probands (8/2,027) and 0.24% of neonates who underwent rapid genome sequencing (1/409). Parental carrier frequency in the birth defects cohort was 0.269% (5/1,858), consistent with previously reported estimates in similarly ascertained families, and 0.098% (3/3,040) in the UDN cohort. All six translocations identified in the birth defects cohort were orthogonally validated using Sanger sequencing, confirming pipeline specificity. Approximately 31% of translocations across the three affected cohorts disrupted a known disease gene at the breakpoint, establishing or refining a molecular diagnosis in five cases previously unresolved. The single translocation identified in the newborn cohort disrupted no disease gene and was classified as a variant of uncertain significance.