Natalie Telis

Background: The Helix Research Network ^TM program is a large population genomics initiative that screens an all-comers population of patients for CDC Tier 1 genetic conditions, including familial hypercholesterolemia (FH). We evaluated changes in clinical management and LDL cholesterol (LDL-C) levels among patients we identified to have FH. Methods: Participants across eight U.S. health systems provided samples that underwent clinical-grade exome sequencing. Individuals with a positive screening result for a Tier 1 condition were offered no-cost genetic counseling through their health system. Using medication and laboratory testing records, we evaluated changes in patients’ lipid-lowering therapies and LDL-C levels. Results: Among 211,263 adults enrolled between 2017-2024, 1,020 (∼1/207) had a pathogenic FH variant in LDLR (72%), APOB (27%), or PCSK9 (1%). Of the 453 with retrospective and prospective electronic health record (EHR) data available (mean of 13.6 and 2.3 years, respectively), 85% lacked a prior clinical FH diagnosis. Overall, 33% received new/modified lipid-lowering therapy within the first year, but this proportion was higher in those with a newly documented FH diagnosis code (55% vs 18% for those without documentation, p<0.001). Patients with new/modified therapies had a mean LDL-C reduction of 59 mg/dL, compared to 18 mg/dL in patients with no therapeutic change (difference=41 mg/dL, p<0.001). Conclusions: Identification of patients with FH likely led to improvements in clinical management and LDL-C levels. EHR documentation of the diagnosis code was associated with greater likelihood of therapeutic modifications which, in turn, were associated with larger LDL-C reductions. Findings underscore the powerful potential of population genomic screening for promoting optimal lipid management in individuals with FH.

Objective: In spite of evidence and recommendations reflecting the importance of pharmacogenomic testing, most prescriptions are still given without testing. We demonstrate the real world implications of the use of testing and evaluate adverse events and outcomes in individuals who did not receive pharmacogenomic testing for clopidogrel. Methods: We analyzed ~100K individuals with paired EHR and exome sequencing data from population health studies administered at multiple medical centers using the Helix Exome+ ^® assay. We inferred clopidogrel dosage by processing the prescription with an LLM. We identified all instances of individuals with at least one prescription that is not in concordance with their CYP2C19 genotype. Lastly, we identify instances of thrombosis using a comprehensive codeset based on ICD9, ICD10, and SNOMED terms. Results: We identified 16,140 prescriptions of clopidogrel given to 3,853 participants. We found that 29% of these individuals have a mismatch between the recommended clopidogrel dosage guideline based on their CYP2C19 genotype and their actual prescribed daily dosage. 25% of poor metabolizers experienced thrombosis, with 40% occurring within 2 months of treatment. Poor and intermediate metabolizers receiving clopidogrel are much more likely to experience thrombosis and myocardial infarction (binomial p- value = 0.001). Conclusions: We estimate a 38% excess of adverse events occur in poor and intermediate metabolizers relative to normal and rapid metabolizers. The lack of testing may be responsible for 1 thrombosis event per every ~30 people prescribed clopidogrel.

Background: Lipoprotein(a) (Lp(a)) levels are a largely genetically determined and often an unmeasured predictor of future Atherosclerotic Cardiovascular Disease (ASCVD). With the increased use of exome sequencing in the clinical setting, there is opportunity to identify patients who have a high chance of having elevated Lp(a) and are therefore at risk of ASCVD. However, accurate genetic predictors of Lp(a) are challenging to design. In addition to single nucleotide variants (SNVs), which are often summarized as a combined genetic risk score, Lp(a) levels are significantly impacted by copy number variation in repeats of the kringle IV subtype 2 domain (KIV-2), which are challenging to quantify. KIV-2 copy numbers are highly variable across populations, and understanding their impact on Lp(a) levels is important to creating an equitable and reliable genetic predictor of Lp(a)-driven cardiovascular risk for all individuals. Methods: We develop a novel method to quantify individuals’ total number of KIV-2 repeats from exome data, validate this quantification against measured Lp(a) levels, and then use this method, combined with a SNV-based genetic risk score, to genotype an entire all-comers cohort of individuals from health systems across the United States (Helix Research Network; N = 76,147) for an estimated Lp(a) level. Results: Our combined genotyping strategy improved prediction of those with clinically-elevated Lp(a) measurements across the genetically diverse cohort, especially for individuals not genetically similar to European reference populations, where GRS-based estimates fall short (r ² = 0.04 for GRS, r ² = 0.34 KIV2+GRS in non-European). Importantly, high combined genetic risk of high Lp(a) genotypes are significantly associated with earlier onset and increased incidence in ASCVD, compared to average and low combined genetic risk genotypes in a retrospective analysis of atherosclerotic diagnoses derived from electronic health records (EHRs). This holds in the cohort at large (CAD HRs=1.29, 1.58), in the European subcohort (HRs=1.30,1.61) as well as at trending levels of significance in individuals not genetically similar to Europeans (HRs=1.22,1.31). In addition, high combined genetic risk for high Lp(a) genotypes are at least 2-fold enriched amongst individuals with ASCVD diagnosis despite a lack of EHR-based evidence of traditional risk factors for cardiovascular disease. Conclusions: Our study demonstrates that genetically predicted Lp(a) levels, incorporating both SNV and our novel KIV-2 repeat estimate, may be a practical method to predict clinically elevated Lp(a). Supporting this, individuals with high combined genetic risk for high Lp(a) have an increased risk for ASCVD, as evidenced across data from seven US-based health systems.

Background: Individual weight loss response to the GLP-1 receptor agonist semaglutide varies considerably, with many possible contributing factors. Leveraging multiple clinico-genomic cohorts, we analyzed differences in weight loss trajectories according to patient characteristics, including a polygenic score (PGS) and metabolic risk factors, in semaglutide initiators with BMI ≥27 kg/m ² . Methods: This longitudinal study utilized clinical-grade exome sequencing and electronic health record data from six U.S. cohorts within the Helix Research Network (n=134,806). A BMI PGS was calculated using 26,941 variants. Twelve-month weight loss trajectories were modeled using mixed effects models, and associations with demographics, PGS, comorbidities, medications, and laboratory results were evaluated. Findings: Among 1,923 semaglutide users, the mean pretreatment BMI was 38.4 kg/m ² . For those on doses ≥1.7 mg, the mean body weight reduction was 7.3% at 6 months and 9.9% at 12 months. Over 12 months, low PGS was associated with an adjusted 1.5% and 1.8% additional weight loss compared to intermediate and high PGS, respectively (both p<0.01). Male sex, type 2 diabetes, hypertension, obstructive sleep apnea, and non-alcoholic fatty liver disease were each associated with 1.2%-1.9% less weight loss (all p<0.05). In type 2 diabetes, each 1%-increase in pretreatment hemoglobin A1c was associated with 0.6% less weight loss (p=0.0019). Interpretation: Among adults with overweight or obesity, a lower genetic predisposition to obesity is linked to greater weight loss on semaglutide. Additionally, metabolic health significantly impacts the drug’s effectiveness. These findings underscore the importance of precision medicine in obesity management. Funding: Renown Health Foundation. Nevada Governor’s Office of Economic Development. HealthPartners.

Increased spinal curvature is one of the most recognizable aging traits in the human population. However, despite high prevalence, the etiology of this condition remains poorly understood. To gain better insight into the physiological, biochemical, and genetic risk factors involved, we developed a novel machine learning method to automatically derive thoracic kyphosis and lumbar lordosis angles from dual-energy X-ray absorptiometry (DXA) scans in the UK Biobank Imaging cohort. In 41,212 participants, we find that on average males and females gain 2.42° kyphotic and 1.48° lordotic angle per decade of life. Increased spinal curvature was strongly associated with decreased muscle mass and bone mineral density. Adiposity had opposing associations, with decreased kyphosis and increased lordosis. To gain further insight into the molecular mechanisms involved, we carried out a genome-wide association study and identified several risk loci associated with both traits. Using Mendelian randomization, we further show that genes fundamental to the maintenance of musculoskeletal function (COL11A1, PTHLH, ETFA, TWIST1) and cellular homeostasis such as RNA transcription and DNA repair (RAD9A, MMS22L, HIF1A, RAB28) are likely involved in increased spinal curvature.

Natural HbA1c levels in GCK -MODY patients often sit above the diagnostic threshold for type 2 diabetes (T2D). Standard treatments to lower HbA1c levels are ineffective in these individuals, yet in case studies to date, GCK -MODY patients often evade secondary T2D complications. Given these deviations from a more typical T2D disease course, genetic screening of GCK may be clinically useful, but population studies are needed to more precisely quantify T2D-related outcomes in GCK variant carriers. Using a state-of-the-art variant interpretation strategy based on glucose elevations, we genotyped all individuals in two real-world cohorts (n~535,000) for GCK risk variants and examined rates of T2D and T2D-complications from seven disease categories. We identified 439 individuals with GCK variants predicted to increase glucose (~1/1200). Aligning with their glucose elevations, GCK -MODY variant carriers were 12x as likely, and all other GCK risk carriers 4x as likely, to receive a T2D diagnosis, compared to non- GCK carriers. Surprisingly, GCK risk carriers with T2D develop a range of T2D-related complications at rates comparable to non- GCK T2D patients. Although the penetrance for secondary complications is lower than that for glucose elevations, GCK risk carriers remain at elevated risk of T2D and secondary complications.

Withdrawal Statement: This version of the manuscript and associated files have been removed owing to concerns of data privacy.

Traces of Neandertal and Denisovan DNA persist in the modern human gene pool, but have been systematically purged by natural selection from genes and other functionally important regions. This implies that many archaic alleles harmed the fitness of hybrid individuals, but the nature of this harm is poorly understood. Here, we show that enhancers contain less Neandertal and Denisovan variation than expected given the background selection they experience, suggesting that selection acted to purge these regions of archaic alleles that disrupted their gene regulatory functions. We infer that selection acted mainly on young archaic variation that arose in Neandertals or Denisovans shortly before their contact with humans; enhancers are not depleted of older variants found in both archaic species. Some types of enhancer appear to have tolerated introgression better than others; compared with tissue-specific enhancers, pleiotropic enhancers show stronger depletion of archaic single-nucleotide polymorphisms. To some extent, evolutionary constraint is predictive of introgression depletion, but certain tissues’ enhancers are more depleted of Neandertal and Denisovan alleles than expected given their comparative tolerance to new mutations. Foetal brain and muscle are the tissues whose enhancers show the strongest depletion of archaic alleles, but only brain enhancers show evidence of unusually stringent purifying selection. We conclude that epistatic incompatibilities between human and archaic alleles are needed to explain the degree of archaic variant depletion from foetal muscle enhancers, perhaps due to divergent selection for higher muscle mass in archaic hominins compared with humans.