23andMe (United States)

Pinpointing causal genes at genome-wide association study (GWAS) loci remains a major bottleneck. Existing literature-mining approaches are often limited in accuracy and scalability. We show that large language models (LLMs) can accurately prioritize likely causal genes at GWAS loci. We systematically evaluated several widely available general-purpose LLMs against benchmark datasets of high-confidence causal genes, including a unique set from 23 unpublished GWAS. Our results demonstrate that LLMs outperform or match current state-of-the-art methods and, crucially, exhibit robust performance on novel loci not previously linked to traits, underscoring their generalizability. Moreover, when integrated with existing methods, LLMs substantially enhance overall performance. This work establishes LLMs as an accurate, scalable, and broadly generalizable approach to accelerate causal gene identification in complex traits.

Abstract Background Polygenic risk scores (PRS) estimate an individual’s germline genetic predisposition to a quantitative trait and/or risk of disease. Several PRS have been developed for cancer risk with the goal of improved risk screening. Here, we sought to establish whether PRS for cancer risk and other common traits may influence survival for patients with cancer. Methods We conducted a PRS survival analysis using 23,770 cancer patients of European ancestry from the Dana-Farber Cancer Institute Profile cohort. Results We identified an association between PRS for breast cancer risk and longer patient survival (HR = 0.89 (95% CI: 0.84–0.95), p = 1.50 × 10–4, < 5% FDR), implying that individuals at high genetic risk had better outcomes. High PRS individuals were also significantly less likely to harbor somatic TP53 mutations, consistent with having less aggressive tumors. This association persisted when including tumor grade and became more protective when restricting to ER-negative tumors (HR = 0.78 (95% CI: 0.68–0.89), p = 1.69 × 10–4). Potential confounders such as hormone receptor status, age, grade, stage, and ER-targeted therapy did not fully explain this association, nor was there statistical evidence of index event bias at individual variants. We did not observe significant associations between cancer risk and survival for other cancers, suggesting that this mechanism may be largely unique to breast cancer. However, we did observe associations between shorter survival and type 2 diabetes, bipolar, and pancreatitis PRS (1% FDR). Conclusions These findings suggest that higher germline risk may predispose individuals to less aggressive breast cancer tumors and provide novel insights into breast cancer development and prognosis.

Completion of the Human Genome Project prompted predictions that genomics would transform medicine, including through genomic screening that identifies potentially medically actionable findings that could prevent disease, detect it earlier, or treat it better. However, genomic screening remains anchored in research and largely unavailable as part of routine care. To summarize 11 years of experience with genomic screening and explore the landscape of genomic screening efforts. This cohort study was based in Geisinger's MyCode Community Health Initiative, a genomic screening program in a rural Pennsylvania health care system in which patient-participants exomes are analyzed. Genomic screen-positive rates were evaluated and stratified by condition type (cancer, cardiovascular, other) and US Centers for Disease Control and Prevention (CDC) Tier 1 designation. The proportion of participants previously unaware of their genomic result was assessed. Other large-scale population-based genomic screening efforts with genomic results disclosure were compiled from public resources. A total of 354 957 patients participated in Geisinger's genomic screening program (median [IQR] age, 54 [36-69] years; 194 037 [59.7%] assigned female sex at birth). As of June 2024, 175 500 participants had exome sequencing available for analysis, and 5934 participants (3.4%) had a pathogenic variant in 81 genes known to increase risk for disease. Between 2013 and July 2024, 5119 results were disclosed to 5052 eligible participants, with 2267 (44.2%) associated with risk for cardiovascular disease, 2031 (39.7%) with risk for cancer, and 821 (16.0%) with risk for other conditions. Most results (3040 [59.4%]) were in genes outside of those with a CDC Tier 1 designation. Nearly 90% of participants (4425 [87.6%]) were unaware of their genomic risk prior to disclosure. In a survey of large-scale biobanks with genomic and electronic health record (EHR) data, only 25.0% (6 of 24) disclosed potentially actionable genomic results. In this large, genomics-informed cohort study from a single health system, 1 in 30 participants had a potentially actionable genomic finding. However, nearly 90% were unaware of their risk prior to screening, demonstrating the utility of genomic screening in identifying at-risk individuals. Most large-scale biobanks with genomic and EHR data did not return genomic results with potential medical relevance, missing opportunities to significantly improve genomic risk ascertainment for these individuals and to perform longitudinal studies of clinical and implementation outcomes in diverse settings.

Abstract ImportanceThe lack of information on progression, phenoconversion, and risk of dementia in a large genotyped sample impedes reliable enrichment for early interventional trials in Parkinson’s disease (PD).ObjectiveTo investigate PD penetrance, risk, motor/non-motor phenotypes, andAPOEallele effects inLRRK2G2019S andGBAN370S carriers.DesignObservational longitudinal case-control self-report survey study.SettingA US population-based study cohort enrolled in the 23andMe, Inc. and Fox Insight Genetic Substudy (FIGS) databases.ParticipantsThe total cohort included 7,586,842 participants (n=35,163 PD; 27% of PD cases from FIGS); 8,791LRRK2G2019S carriers (565 with PD), 37,427GBAN370S carriers (524 with PD), 244 dual carriers (37 with PD), and 7.5 million non-carriers (34,037 with PD).Exposure(s)LRRK2G2019S,GBAN370S,APOEE2/E3/E4 alleles and PD polygenic risk scores (PRS).Main Outcome(s) and Measure(s)Cumulative incidence of PD was estimated using Kaplan-Meier and accelerated failure time models. Relative odds of developing motor and non-motor symptoms were calculated using logistic regression models according to genetic exposure. Impact of theAPOEalleles was estimated in a dose-dependent analysis.ResultsBy the age of 80 years, the cumulative incidence of PD was 43% for dual carriers, 32% forLRRK2G2019S carriers, 6% forGBAN370S carriers, and 3% for non-carriers. Higher PRS was associated with increased penetrance of the variants and earlier time to PD diagnosis. Motor symptoms were similar inLRRK2G2019S,GBAN370S, and non-carriers with PD.GBAN370S PD was associated with the highest burden of non-motor symptoms, including REM sleep behavior disorder and cognitive/memory deficits, andLRRK2G2019S the lowest.APOEE4 dosage was associated with greater odds of developing hallucinations and cognitive decline in addition to carrier status.Conclusions and RelevanceOur findings support the use of genetic screening—includingLRRK2G2019S,GBAN370S,APOEE4, and PRS—to enrich candidate selection for neuroprotective trials and better define outcome measures based on genetic risk factors.

Dyslexia is a learning difficulty with neurodevelopmental origins, manifesting as reduced accuracy and speed in reading and spelling. It is substantially heritable and frequently co-occurs with other neurodevelopmental conditions, particularly attention deficit-hyperactivity disorder (ADHD). Here, we investigate the genetic structure underlying dyslexia and a range of psychiatric traits using results from genome-wide association studies of dyslexia, ADHD, autism, anorexia nervosa, anxiety, bipolar disorder, major depressive disorder, obsessive compulsive disorder, schizophrenia, and Tourette syndrome. Genomic Structural Equation Modelling (GenomicSEM) showed heightened support for a model consisting of five correlated latent genomic factors described as: F1) compulsive disorders (including obsessive-compulsive disorder, anorexia nervosa, Tourette syndrome), F2) psychotic disorders (including bipolar disorder, schizophrenia), F3) internalising disorders (including anxiety disorder, major depressive disorder), F4) neurodevelopmental traits (including autism, ADHD), and F5) attention and learning difficulties (including ADHD, dyslexia). ADHD loaded more strongly on the attention and learning difficulties latent factor (F5) than on the neurodevelopmental traits latent factor (F4). The attention and learning difficulties latent factor (F5) was positively correlated with internalising disorders (.40), neurodevelopmental traits (.25) and psychotic disorders (.17) latent factors, and negatively correlated with the compulsive disorders (–.16) latent factor. These factor correlations are mirrored in genetic correlations observed between the attention and learning difficulties latent factor and other cognitive, psychological and wellbeing traits. We further investigated genetic variants underlying both dyslexia and ADHD, which implicated 49 loci (40 not previously found in GWAS of the individual traits) mapping to 174 genes (121 not found in GWAS of individual traits) as potential pleiotropic variants. Our study confirms the increased genetic relation between dyslexia and ADHD versus other psychiatric traits and uncovers novel pleiotropic variants affecting both traits. In future, analyses including additional co-occurring traits such as dyscalculia and dyspraxia will allow a clearer definition of the attention and learning difficulties latent factor, yielding further insights into factor structure and pleiotropic effects.

SummaryCannabis is one of the most widely used drugs globally. Decriminalization of cannabis is further increasing cannabis consumption. We performed genome-wide association studies (GWASs) of lifetime (N=131,895) and frequency (N=73,374) of cannabis use. Lifetime cannabis use GWAS identified two loci, one near CADM2 (rs11922956, p=2.40E-11) and another near GRM3 (rs12673181, p=6.90E-09). Frequency of use GWAS identified one locus near CADM2 (rs4856591, p=8.10E-09; r2=0.76 with rs11922956). Both traits were heritable and genetically correlated with previous GWASs of lifetime use and cannabis use disorder (CUD), as well as other substance use and cognitive traits. Polygenic scores (PGSs) for lifetime and frequency of cannabis use associated cannabis use phenotypes in AllofUs participants. Phenome-wide association study of lifetime cannabis use PGS in a hospital cohort replicated associations with substance use and mood disorders, and uncovered associations with celiac and infectious diseases. This work demonstrates the value of GWASs of CUD transition risk factors.

Fox Insight is an online, longitudinal study of over 54,000 people with and without Parkinson’s disease, facilitating discovery, validation, and reproducibility in Parkinson’s disease research. The study administers routine longitudinal assessments, one-time questionnaires on an array of topics such as environmental exposure or COVID-19, plus genetic and microbiome data collection. Researchers can explore and download patient-reported outcomes data and Parkinson’s disease related genetic variants upon completing a Data Use Agreement. The full genetic data set, including approximately 650,000 single nucleotide polymorphisms for over 10,000 participants, and the microbiome data set for over 650 participants, can be requested with a heightened level of access. Since the first Fox Insight data descriptor was published in 2020, the data captured has been extended significantly, so this paper supersedes the previous one. Since then, the number of participants has increased by more than 20,000; an additional 1,747,729 surveys were completed; 130 gigabytes of genetic data were released; responses from 16 new one-time surveys were collected; and, data from one additional sub-study was made available.

The National Health and Nutrition Examination Survey provides comprehensive data on demographics, sociology, health and nutrition. Conducted in 2-year cycles since 1999, most of its data are publicly accessible, making it pivotal for research areas like studying social determinants of health or tracking trends in health metrics such as obesity or diabetes. Assembling the data and analyzing it presents a number of technical and analytic challenges. This paper introduces the nhanesA R package, which is designed to assist researchers in data retrieval and analysis and to enable the sharing and extension of prior research efforts. We believe that fostering community-driven activity in data reproducibility and sharing of analytic methods will greatly benefit the scientific community and propel scientific advancements. Database URL: https://github.com/cjendres1/nhanes.

ABSTRACTA substantial proportion of acute SARS-CoV2 infection cases exhibit gastrointestinal symptoms, yet the genetic determinants of these extrapulmonary manifestations are poorly understood. Using survey data from 239,866 individuals who tested positively for SARS-CoV2, we conducted a multi-ancestry GWAS of 80,289 cases of diarrhea occurring during acute COVID-19 infection (33.5%). Six loci (CYP7A1, LZFTL1- -CCR9, TEME182, NALCN, LFNG, GCKR) met genome-wide significance in a trans-ancestral analysis. The top significant GWAS hit mapped to theCYP7A1locus, which plays an etiologic role in bile acid metabolism and is in high LD (r2= 0.93) with theSDCBPgene, which was previously implicated in antigen processing and presentation in the COVID-19 context. Another association was observed with variants in theLZTFL1–CCR9region, which is a known locus for COVID-19 susceptibility and severity. PheWAS showed a shared association across three of the six SNPs with irritable bowel syndrome (IBS) and its subtypes. Mendelian randomization showed that genetic liability to IBS-diarrhea increased (OR=1.40,95%,CI[1.33-1.47]), and liability to IBS-constipation decreased (OR=0.86, 95%CI[0.79-0.94]) the relative odds of experiencing COVID-19+ diarrhea. Our genetic findings provide etiological insights into the extrapulmonary manifestations of acute SARS-CoV2 infection.